While pkgconfig is supposed to be a distribution and version neutral way
of detecting packages, it doesn't always work this way. The sd_notify()
manual page documents that sd_notify is available via the libsystemd
package, but on centos 7.0 it is only available via the libsystemd-daemon
package (on centos 7.1+ it works as expected).
Fix by allowing for alternate version of package names, testing each one
until a match is found.
Fixes#879.
Message-Id: <1454858862-5239-1-git-send-email-avi@scylladb.com>
(cherry picked from commit 8b0a26f06d)
" - Added waiting for async cleanup on clean shutdown
- Crash in the middle of sstable removal doesn't leave system in a non-bootable state"
(cherry picked from commit f3ca597a01)
Generated sstables may imply either fully or partially written.
Compaction is interrupted if it was deriberately asked to stop (stop API)
or it was forced to do so in event of a failure, ex: out of disk space.
There is a need to explicitly delete sstables generated by a compaction
that was interrupted. Otherwise, such sstables will waste disk space and
even worsen read performance, which degrades as number of generations
to look at increases.
Fixes#852.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <49212dbf485598ae839c8e174e28299f7127f63e.1453912119.git.raphaelsc@scylladb.com>
storage_service::get_local_ranges returns sorted ranges, which are
not overlapping nor wrap-around. As a result, there is no need for
the consumer to do anything.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
supervisor_notify() calls periodically, to log message on systemd.
So raise(SIGSTOP) will called multiple times, upstart doesn't expected that.
We need to call it just one time.
Fixes#846
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Upstart job able to specify ulimit like systemd, so drop ubuntu's scylla_run and merge with redhat one.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Not all the idls are used by the messaging service, this patch removes
the auto-generated single include file that holds all the files and
replaes it with individual include of the generated fiels.
The patch does the following:
* It removes from the auto-generated inc file and clean the configure.py
from it.
* It places an explicit include for each generated file in
messaging_serivce.
* It add dependency of the generated code in the idl-compiler, so a
change in the compiler will trigger recreation of the generated files.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Message-Id: <1453900241-13053-1-git-send-email-amnon@scylladb.com>
I saw the following Boost format string related warning during commitlog
replay:
INFO [shard 0] commitlog_replayer - Replaying node3/commitlog/CommitLog-1-72057594289748293.log, node3/commitlog/CommitLog-1-90071992799230277.log, node3/commitlog/CommitLog-1-108086391308712261.log, node3/commitlog/CommitLog-1-251820357.log, node3/commitlog/CommitLog-1-54043195780266309.log, node3/commitlog/CommitLog-1-36028797270784325.log, node3/commitlog/CommitLog-1-126100789818194245.log, node3/commitlog/CommitLog-1-18014398761302341.log, node3/commitlog/CommitLog-1-126100789818194246.log, node3/commitlog/CommitLog-1-251820358.log, node3/commitlog/CommitLog-1-18014398761302342.log, node3/commitlog/CommitLog-1-36028797270784326.log, node3/commitlog/CommitLog-1-54043195780266310.log, node3/commitlog/CommitLog-1-72057594289748294.log, node3/commitlog/CommitLog-1-90071992799230278.log, node3/commitlog/CommitLog-1-108086391308712262.log
WARN [shard 0] commitlog_replayer - error replaying: boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<boost::io::too_many_args> > (boost::too_many_args: format-string referred to less arguments than were passed)
While inspecting the code, I noticed that one of the error loggers is
missing an argument. As I don't know how the original failure triggered,
I wasn't able to verify that that was the only one, though.
Message-Id: <1453893301-23128-1-git-send-email-penberg@scylladb.com>
Any stream, no matter initialized by us or initialized by a peer node,
can send and receive data. We should audit incoming/outgoing bytes in
the all streams.
The main motivation behind this change is to make get_ranges() easier for
consumers to work with the returned ranges, e.g. binary search to find a
range in which a token is contained. In addition, a wrap-around range
introduces corner cases, so we should avoid it altogether.
Suppose that a node owns three tokens: -5, 6, 8
get_ranges() would return the following ranges:
(8, -5], (-5, 6], (6, 8]
get_ranges() will now return the following ranges:
(-inf, -5], (-5, 6], (6, 8], (8, +inf)
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <4bda1428d1ebbe7c8af25aa65119edc5b97bc2eb.1453827605.git.raphaelsc@scylladb.com>
We register engine().at_exit() callbacks when we initialize the services. We
do not really call the callbacks at the moment due to #293.
It is pretty hard to see the whole picture in which order the services
are shutdown. Instead of for each services to register a at_exit()
callbacks, I proposal to have a single at_exit() callback which do the
shutdown for all the services. In cassandra, the shutdown work is done
in storage_service::drain_on_shutdown callbacks.
In this patch, the drain_on_shutdown is executed during shutdown.
As a result, the proper gossip shutdown is executed and fixes#790.
With this patch, when Ctrl-C on a node, it looks like:
INFO [shard 0] storage_service - Drain on shutdown: starts
INFO [shard 0] gossip - Announcing shutdown
INFO [shard 0] storage_service - Node 127.0.0.1 state jump to normal
INFO [shard 0] storage_service - Drain on shutdown: stop_gossiping done
INFO [shard 0] storage_service - CQL server stopped
INFO [shard 0] storage_service - Drain on shutdown: shutdown rpc and cql server done
INFO [shard 0] storage_service - Drain on shutdown: shutdown messaging_service done
INFO [shard 0] storage_service - Drain on shutdown: flush column_families done
INFO [shard 0] storage_service - Drain on shutdown: shutdown commitlog done
INFO [shard 0] storage_service - Drain on shutdown: done
Time a node waits after sending gossip shutdown message in milliseconds.
Reduces ./cql_query_test execution time
from
real 2m24.272s
user 0m8.339s
sys 0m10.556s
to
real 1m17.765s
user 0m3.698s
sys 0m11.578
row_cache::update() does not explicitly invalidate the entries it failed
to update in case of a failure. This could lead to inconsistency between
row cache and sstables.
In paractice that's not a problem because before row_cache::update()
fails it will cause all entries in the cache to be invalidated during
memory reclaim, but it's better to be safe and explicitly remove entries
that should be updated but it was not possible to do so.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Message-Id: <1453829681-29239-1-git-send-email-pdziepak@scylladb.com>
The API needs to be available at an early stage of the initialization,
on the other hand not all the specific APIs are available at that time.
This patch breaks the API initialization into stages, in each stage
additional commands will be available.
While setting that the api header files was broken into api_init.hh that
is relevent to the main and to api.hh which holds the different
api helper functions.
Fixes#754
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Message-Id: <1453822331-16729-2-git-send-email-amnon@scylladb.com>
Last series accidently broke batch mode.
With new, fancy, potentitally blocking ways, we need to treat
batch mode differently, since in this case, sync should always
come _after_ alloc-write.
Previous patch caused infinite loop. Broke jenkins.
Message-Id: <1453821077-2385-1-git-send-email-calle@scylladb.com>
Also check closed status in allocate, since alloc queue waiting could
lead to us re-allocating in a segment that gets closed in between
queue enter and us running the continuation.
Message-Id: <1453811471-1858-1-git-send-email-calle@scylladb.com>
invocation of sstable_test "./test.py --name sstable_test --mode
release --jenkins a"
ran ... --log_sink=a.release.sstable_test -c1.boost.xml" which caused
the test to fail "with error code -11" fix that.
In addition boost test printout was bad fix that as well
Signed-off-by: Shlomi Livne <shlomi@scylladb.com>
Message-Id: <3af8c4b55beae673270f5302822d7b9dbba18c0f.1453809032.git.shlomi@scylladb.com>
"Adds flush + write thresholds/limits that, when reached, causes
operations to wait before being issued.
Write ops waiting also causes further allocations to queue up,
i.e. limiting throughput.
Adds getters for some useful "backlog" measurements:
* Pending (ongoing) writes/flush
* Pending (queued, wating) allocations
* Num times write/flush threshold has been exceeded (i.e. waits occured)
* Finished, dirty segments
* Unused (preallocated) segments"
* seastar 97f418a...bdb273a (6):
> rpc: alias rpc::type to boost::type
> Fix warning_supported to properly work with Clang
> rpc: change 'overflow' to 'underflow' in input stream processing
> rpc: log an error that caused connection to be closed.
> rpc: clarify deserialization error message
> rpc: do not append new line in a logger
Configured on start (for now - and dummy values at that).
When shard write/flush count reaches limit, and incoming ops will queue
until previous ones finish.
Consequently, if an allocation op forces a write, which blocks, any
other incoming allocations will also queue up to provide back pressure.
"After the patch, all of our relevant I/O is placed on a specific priority class.
The ones which are not are left into the Seastar's default priority, which will
effectively work as an idle class.
Examples of such I/O are commitlog replay and initial SSTable loading. Since they
will happen during initialization, they will run uncontended, and do not justify
having a class on their own."
It is wrong to get a stream plan id like below:
utils::UUID plan_id = gms::get_local_gossiper().get_host_id(ep);
We should look at all stream_sessions with the peer in question.
There are only two messages: prepare_message and outgoing_file_message.
Actually only the prepare_message is the message we send on wire.
Flatten the namespace.
After this patch, our I/O operations will be tagged into a specific priority class.
The available classes are 5, and were defined in the previous patch:
1) memtable flush
2) commitlog writes
3) streaming mutation
4) SSTable compaction
5) CQL query
Signed-off-by: Glauber Costa <glauber@scylladb.com>
After the introduction of the Fair I/O Queueing mechanism in Seastar,
it is possible to add requests to a specific priority class, that will
end up being serviced fairly.
This patch introduces a Priority Manager service, that manages the priority
each class of request will get. At this moment, having a class for that may
sound like an overkill. However, the most interesting feature of the Fair I/O
queue comes from being able to adjust the priorities dynamically as workloads
changes: so we will benefit from having them all in the same place.
This is designed to behave like one of our services, with the exception that
it won't use the distributed interface. This is mainly because there is no
reason to introduce that complexity at this point - since we can do thread local
registration as we have been doing in Seastar, and because that would require us
to change most of our tests to start a new service.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
SSTables already have a priority argument wired to their read path. However,
most of our reads do not call that interface directly, but employ the services
of a mutation reader instead.
Some of those readers will be used to read through a mutation_source, and those
have to patched as well.
Right now, whenever we need to pass a class, we pass Seastar's default priority
class.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
All the SSTable read path can now take an io_priority. The public functions will
take a default parameter which is Seastar's default priority.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
All variants of write_component now take an io_priority. The public
interfaces are by default set to Seastar's default priority.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
The only user for the default size is data_read, sitting at row.cc.
That reader wants to read and process a chunk all at once. So there's
really no reason to use the default buffer size - except that this code
is old.
We should do as we do in other single-key / single-range readers and
try to read all at once if possible, by looking at the size we received
as a parameter. Cleaning up the data_stream_at interface then comes as
a nice side effect.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Its definition as a lambda function is inconvenient, because it does not allow
us to use default values for parameters.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Its definition as a lambda function is inconvenient, because it does not allow
us to use default values for parameters.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
In the existing code, when we fail to reach one of the replicas of some
range being repaired, we would give up, and not continue to repair the
living replicas of this range. The thinking behind this was since the
repair should be considered failed anyway, there's no point in trying
to do a half-job better.
However, in a discussion I had with Shlomi, he raised the following
alternative thinking, which convinced me: In a large cluster, having
one node or another temporarily dead has a high probability. In that
case, even if the if the repair is doomed to be considered "failed",
we want it at least to do as much as it possibly can to repair the
data on the living part of the cluster. This is what this patch does:
If we can only reach some of the replicas of a given range, the repair
will be considered failed (as before), but we will still repair the
reachable replicas of this range, if they have different checksums.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Message-Id: <1453724443-29320-1-git-send-email-nyh@scylladb.com>
Unlike streaming in c*, scylla does not need to open tcp connections in
streaming service for both incoming and outgoing messages, seastar::rpc
does the work. There is no need for a standalone stream_init_message
message in the streaming negotiation stage, we can merge the
stream_init_message into stream_prepare_message.
The helper is used only once in init_sending_side and in
init_receiving_side we do not use create_and_register to create
stream_result_future. Kill the trivial helper to make the code more
consistent.
In addition, rename variables "future" and "f" to sr (streaming_result).
So we have:
- init_sending_side
called when the node initiates a stream_session
- init_receiving_side
called when the node is a receiver of a stream_session initiated by a peer
In scylla, in each stream_coordinator, there will be only one
stream_session for each remote peer. Drop the code supporting multiple
stream_sessions in host_streaming_data.
We now have
shared_ptr<stream_session> _stream_session
instead of
std::map<int, shared_ptr<stream_session>> _stream_sessions
- from
We can get it form the rpc::client_info
- session_index
There will always be one session in stream_coordinator::host_streaming_data with a peer.
- is_for_outgoing
In cassandra, it initiates two tcp connections, one for incoming stream and one for outgoing stream.
logger.debug("[Stream #{}] Sending stream init for incoming stream", session.planId());
logger.debug("[Stream #{}] Sending stream init for outgoing stream", session.planId());
In scylla, it only initiates one "connection" for sending, the peer initiates another "connection" for receiving.
So, is_for_outgoing will also be true in scylla, we can drop it.
- keep_ss_table_level
In scylla, again, we stream mutations instead of sstable file. It is
not relevant to us.
- int connections_per_host
Scylla does not create connections per stream_session, instead it uses
rpc, thus connections_per_host is not relevant to scylla.
- bool keep_ss_table_level
- int repaired_at
Scylla does not stream sstable files. They are not relevant to scylla.
"This series:
- Add more debug info to stream session
- Fail session if we fail to send COMPLETE_MESSAGE
- Handle message retry logic for verbs used by streaming
See commit log for details."
"The series do the following:
It adds the code generation
Perform the needed changes in the current classes so each would have getter for
each of its serializable value and a constructor from the serialized values.
It adds a schema definition that cover gossip_diget_ack
It changes the messaging_service to use the generated code.
An overall explanation of the solution with a description of the schema IDL can
be found on the wiki page:
https://github.com/scylladb/scylla/wiki/Serializer-Deserializer-Code-generation
"
Serializer requires class to be defined, so it has to be in .h file. It
also does not support nested types yet, so move it outside of containing
class.
python3 needs to install pyparsing excplicitely. This adds the
installation of python3-pyparsing to the require dependencies in the
README.md
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
the serializer
This patch adds a specific template initialization so that the rpc would
use the serializer and deserializer that are auto-generated.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
This patch adds the serializer and serializer_imple files. They holds
the functions that are not auto generated: primitives and templates (map
and vector) It also holds the include to the auto-generated code.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
This patch adds rules and the idl schema to configure, which will call
the code generation to create the serialization and deserialization
functions.
There is also a rule to create the header file that include the auto
generated header files.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
This is a definition example for gossip_digest_ack with all its sub
classes.
It can be used by the code generator to create the serializer and
deserializer functions.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
inet_address uses uint32_t to store the ip address, but its constructor
is int32_t.
So this patch adds such a constructor.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
This patch contains two changes, it make the constructor with parameters
public. And it removes the dependency in messaging_service.hh from the
header file by moving some of the code to the .cc file.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
The code generation takes a schema file and create two files from it,
one with a dist.hh extension containing the forward declarations and a second
with dist.impl.hh with the actual implementation.
Because the rpc uses templating for the input and output streams. The
generated functions are templates.
For each class, struct or enum, two functions are created:
serialize - that gets the output buffer as template parameter and
serialize the object to it. There must be a public way to get to each of
the parameters in the class (either a getter or the parameter should be
public)
deserialize - that gets an input buffer, and return the deserialize
object (and by reference the number of char it read).
To create the return object, the class must have a public constructor
with all of its parameters.
The solution description can be found here:
https://github.com/scylladb/scylla/wiki/Serializer-Deserializer-Code-generation
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
When a verb timeout, if we resend the message again, the peer could receive
the message more than once. This would confuse the receiver. Currently, only
the streaming code use the retry logic.
- In case of rpc:timeout_error:
Instead of doing timeout in a relatively short time and resending a few
times, we make the timeout big enough and let the tcp to do the resend.
Thus, we can avoid resending the message more than once, of course, the
receiver would not receive the message more than once.
- In case of rpc::closed_error:
There are two cases:
1) Failing to establish a connection.
For instance, the peer is down. It is safe to resend since we know for
sure the receiver hasn't received the message yet.
2) The connection is established.
We can not figure out if the remote peer have received the message
already or not upon receiving the rpc::closed_error exception.
Currently, we still sleep & resend the message again, so the receiver
might receive the message more than once. We do not have better choice
in this case, if we want the resend to recover the sending error due to
temporary network issue, since failing the whole stream_session due to
failing to send a single message is not wise.
NOTE: If the duplicated message is received when the stream_session is done,
it will be ignored since it can not find the stream_manager anymore.
For message like, STREAM_MUTATION, it is ok to receive twice (we apply the
mutation twice).
TODO: For other messages which uses the retry logic, we need
to make sure it is ok to receive more than once.
- Add debug for the peer address info
- Add debug in stream_transfer_task and stream_receive_task
- Add debug when cancel the keep_alive timer
- Add debug for has_active_sessions in stream_result_future::maybe_complete
Compaction manager was initially created at utils because it was
more generic, and wasn't only intended for compaction.
It was more like a task handler based on futures, but now it's
only intended to manage compaction tasks, and thus should be
moved elsewhere. /sstables is where compaction code is located.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
"This series moves the "backup" logic into the sstable::write_components()
methods, adds a support for enabling backup for sstables flushed in the
compaction flow (in addition to a regular flushing flow which had this support
already) and enables the "incremental_backups" configuration option."
I fixed up a merge conflict with commit 5e953b5 ("Merge "Add support to
stop ongoing compaction" from Raphael").
"stop compaction is about temporarily interrupting all ongoing compaction
of a given type.
That will also be needed for 'nodetool stop <compaction_type>'.
The test was about starting scylla, stressing it, stopping compaction using
the API and checking that scylla was able to recover.
Scylla will print a message as follow for each compaction that was stopped:
ERROR [shard 0] compaction_manager - compaction failed: read exception:
std::runtime_error (Compaction for keyspace1/standard1 was deliberately stopped.)
INFO [shard 0] compaction_manager - compaction task handler sleeping for 20 seconds"
Enable the incremental_backups/--incremental-backups option.
When enabled there will be a hard link created in the
<column family directory>/backup directory for every flushed
sstable.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Enable incremental backup when sstables are flushed if
incremental backup has been requested.
It has been enabled in the regular flushing flow before but
wasn't in the compaction flow.
This patch enables it in both places and does it using a
backup capability of sstable::write_components() method(s).
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
When 'backup' parameter is TRUE - create backup hard
links for a newly written sstables in <sstable dir>/backups/
subdirectory.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Only difference from previous sleep is that we will
explicitly delete the objects if the process terminates
before tasks are run. I.e. make ASas happier.
Message-Id: <1453295521-29580-1-git-send-email-calle@scylladb.com>
Arguments buffer_size and true were accidently inverted.
GCC wasn't complaning because implicit conversion of bool to
int, and vice-versa, is valid.
However, this conversion is not very safe because we could
accidentaly invert parameters.
This should fix the last problem with sstable_test.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <9478cd266006fdf8a7bd806f1c612ec9d1297c1f.1453301866.git.raphaelsc@scylladb.com>
test_cassandra_hash also sort of expects exceptions. ASas causes false
positives here as well with seastar::thread, so do it with normal cont.
Message-Id: <1453295521-29580-2-git-send-email-calle@scylladb.com>
Since 581271a243 "sstables: ignore data
belonging to dropped columns" we silently drop cells if there is no
column in the current schema that they belong to or their timestamp is
older than the column dropped_at value. Originally this check was
applied to row markers as well which caused them to be always dropped
since there is no column in the schema representing these markers.
This patch makes sure that the check whether colum is alive is performed
only if the cell is not a row marker.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Message-Id: <1453289300-28607-1-git-send-email-pdziepak@scylladb.com>
stop_compaction is implemented by calling stop_compaction() of
compaction manager for each database.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
That's needed for nodetool stop, which is called to stop all ongoing
compaction. The implementation is about informing an ongoing compaction
that it was asked to stop, so the compaction itself will trigger an
exception. Compaction manager will catch this exception and re-schedule
the compaction.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
compaction_info makes more sense because this structure doesn't
only store stats about ongoing compaction. Soon, we will add
information to it about whether or not an user asked to stop the
respective ongoing compaction.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
From Paweł:
These patches contain fixes for date and timestamp types:
- date and timestamp are considered compatible types
- date type is added to abstract_type::parse_type()
When this code was originally written, we used to operate on a generic
output_stream. We created a file output stream, and then moved it into
the generic object.
Many patches and reworks later, we now have a file_writer object, but
that pattern was never reworked.
So in a couple of places we have something like this:
f = file_object acquired by open_file_dma
auto out = file_writer(std::move(f), 4096);
auto w = make_shared<file_writer>(std::move(out));
The last statement is just totally redundant. make_shared can create
an object from its parameters without trouble, so we can just pass
the parameter list directly to it.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <c01801a1fdf37f8ea9a3e5c52cd424e35ba0a80d.1453219530.git.glauber@scylladb.com>
* Match origin log messages
- Demote per-file printouts to "debug" level.
* Print an all-files stat summary for whole replay (begin/summary)
- At info level, like origin
Prompted by dtest that expects origin log output.
Message-Id: <1453216558-18359-1-git-send-email-calle@scylladb.com>
"We need to be able to replay mutations created using older versions of
the table's schema. frozen_mutation can be only read using the version
it was serialized with, and there is no guarantee that the node will
know this version at the time of replay. Currently versions are kept
in-memory so a node forgets all past versions when it restarts. This
was not implemented yet, replay would fail with exception if the
version is unknown."
* Match origin log messages
- Demote per-file printouts to "debug" level.
* Print an all-files stat summary for whole replay (begin/summary)
- At info level, like origin
Prompted by dtest that expects origin log output.
v2:
* Fixed broken + operator
* Use map_reduce instead of easily readable code
We need to be able to replay mutations created using older versions of
the table's schema. frozen_mutation can be only read using the version
it was serialized with, and there is no guarantee that the node will
know this version at the time of replay. Currently versions kept
in-memory so a node forgets all past versions when it restarts.
To solve this, let's store canonical_mutations which, like data in
sstables, can be read using any later schema version of given table.
From Paweł:
"This series contains some more fixes for issues related to alter table,
namely: incorrect parsing of collection information in comparator, missing
schema::_raw._collections in equality check, missing compatibility
information for utf8->blob, ascii->blob and ascii->utf8 casts."
test_password_authenticator_operations causes ASan failures, in a way
that I am 99% sure is fully false positive, caused by a combo of
seastar threads, exception throwing and externals.
In lieu of actually identifying what ASan flaw causes this and
potentially cure it, for now, lets just re-write the test in question
to not use seastar::async, but normal continuation. Less easy to read,
but passes ASan.
Message-Id: <1453205136-10308-1-git-send-email-calle@scylladb.com>
test_password_authenticator_operations causes ASan failures, in a way
that I am 99% sure is fully false positive, caused by a combo of
seastar threads, exception throwing and externals.
In lieu of actually identifying what ASan flaw causes this and
potentially cure it, for now, lets just re-write the test in question
to not use seastar::async, but normal continuation. Less easy to read,
but passes ASan.
When compacting sstable, mutation that doesn't belong to current shard
should be filtered out. Otherwise, mutation would be duplicated in
all shards that share the sstable being compacted.
sstable_test will now run with -c1 because arbitrary keys are chosen
for sstables to be compacted, so test could fail because of mutations
being filtered out.
fixes#527.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Message-Id: <1acc2e8b9c66fb9c0c601b05e3ae4353e514ead5.1453140657.git.raphaelsc@scylladb.com>
Don't demand the messaging_service version to be the same on both
sides of the connection in order to use internal addresses.
Upstream has a similar change for CASSANDRA-6702 in commit a7cae32 ("Fix
ReconnectableSnitch reconnecting to peers during upgrade").
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Message-Id: <1452686729-32629-1-git-send-email-vladz@cloudius-systems.com>
The correct format of collection information in comparator is:
o.a.c.db.m.ColumnToCollection(<name1>:<type1>, <name2>:<type2>, ...)
not:
o.a.c.db.m.ColumnToCollection(<name1>:<type1>),
o.a.c.db.m.ColumnToCollection(<name2>:<type2>) ...
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
The upstream of origin adds the version to the application_state in the
get_endpoints in the failure detector.
In our implementation we return an object to the jmx proxy and the proxy
do the string formatting.
This patch adds the version to the return object which is both useful as
an API and will allow the jmx proxy to add it to its output when we move
forward with the jmx version.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Message-Id: <1448962889-19611-1-git-send-email-amnon@scylladb.com>
Representation format is an implementation detail of
partition_key. Code which compares a value to representation makes
assumptions about key's representation. Compare keys to keys instead.
Message-Id: <1453136316-18125-1-git-send-email-tgrabiec@scylladb.com>
We cannot stop the stream manager because it's accessible via the API
server during shutdown, for example, which can cause a SIGSEGV.
Spotted by ASan.
Message-Id: <1453130811-22540-1-git-send-email-penberg@scylladb.com>
Fix various issues in set_messaging_service() that caused
heap-buffer-overflows when JMX proxy connects to Scylla API:
- Off-by-one error in 'num_verb' definition
- Call to initializer list std::vector constructor variant that caused
the vector to be two elements long.
- Missing verb definitions from the Swagger definition that caused
response vector to be too small.
Spotted by ASan.
Message-Id: <1453125439-16703-1-git-send-email-penberg@scylladb.com>
Wait for batchlog removal before completing a query otherwise batchlog
removal queries may accumulate. Still ignore an error if it happens
since it is not critical, but log it.
Message-Id: <20160118095642.GB6705@scylladb.com>
"This series makes sure that Scylla rejects adding a collections if
its column name is the same as a collection that existed before and
their types are incompatible.
Fixes#782"
Historically the purpose of the metric is to show how much memory is
in standard allocations. After zones were introduced, this would also
include free space in lsa zones, which is almost all memory, and thus
the metric lost its original meaning. This change brings it back to
its original meaning.
Message-Id: <1452865125-4033-1-git-send-email-tgrabiec@scylladb.com>
"This patch is intended to add support to column family cleanup, which will
make 'nodetool cleanup' possible.
Why is this feature needed? Remove irrelevant data from a node that loses part
of its token range to a newly added node."
"This series adds support for multiple schema versions to the commit log.
All segments contain column mappings of all schema versions used by the
mutations contained in the segment, which are necessary in order to be
able to read frozen mutations and upgrade them to the current schema
version."
While MUTATION and MUTATION_DONE are asynchronous by nature (when a MUTATION
completes, it sends a MUTATION_DONE message instead of responding
synchronously), we still want them to be synchronous at the server side
wrt. the RPC server itself. This is because RPC accounts for resources
consumed by the handler only while the handler is executing; if we return
immediately, and let the code execute asynchronously, RPC believes no
resources are consumed and can instantiate more handlers than the shard
has resources for.
Fix by changing the return type of the handlers to future<no_wait_type>
(from a plain no_wait_type), and making that future complete when local
processing is over.
Ref #596.
Message-Id: <1453048967-5286-1-git-send-email-avi@scylladb.com>
Theoretically, one could want to repair a single host *and* all the hosts
in one or more other data centers which don't include this host. However,
Cassandra's "nodetool repair" explicitly does not allow this, and fails if
given a list of data centers (via the "-dc" option) which doesn't include
the host starting the repair. So we need to behave like "nodetool repair"
and fail in this case too.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Message-Id: <1453037016-25775-1-git-send-email-nyh@scylladb.com>
Cassandra disallows adding a column with the same name as a collection
that existed in the past in that table if the types aren't compatible.
To enforce that Scylla needs to keep track of all collections that ever
existed in the column family.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
* seastar a8183c1...e93cd9d (2):
> rpc: make sure we serialize on _resources_available sempahore
> rpc: fix support for handlers returning future<no_wait_type>
Since we switched to use mk-build-deps, it only resolves build-time dependencies.
We also need to install install-time dependencies.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
This patch adds a function which waits for the background cleanup work
which is started from sstable destructors.
We wait for those cleanups on reactor exit so that unit tests don't
leak. This fixes erratic ASAN complaint about memory leak when running
schema_change_test in debug mode:
Indirect leak of 64 byte(s) in 1 object(s) allocated from:
0x7fab24413912 in operator new(unsigned long) (/lib64/libasan.so.2+0x99912)
0x1776aeb in make_unique<continuation<future<T>::then_wrapped(Func&&) [with Func = future<T>::handle_exception(Func&&) [with Func = sstables::sstable::~sstable()::<lambda(auto:52)>; T = {}]::<lambda(auto:5&&)>; Result = future<>; T = {}]::<lambda(auto:2&&)> >, future<T>::then_wrapped(Func&&) [with Func = future<T>::handle_exception(Func&&) [with Func = sstables::sstable::~sstable()::<lambda(auto:52)>; T = {}]::<lambda(auto:5&&)>; Result = future<>; T = {}]::<lambda(auto:2&&)> > /usr/include/c++/5.1.1/bits/unique_ptr.h:765
0x1752b69 in schedule<future<T>::then_wrapped(Func&&) [with Func = future<T>::handle_exception(Func&&) [with Func = sstables::sstable::~sstable()::<lambda(auto:52)>; T = {}]::<lambda(auto:5&&)>; Result = future<>; T = {}]::<lambda(auto:2&&)> > /home/tgrabiec/src/scylla2/seastar/core/future.hh:513
0x1711365 in schedule<future<T>::then_wrapped(Func&&) [with Func = future<T>::handle_exception(Func&&) [with Func = sstables::sstable::~sstable()::<lambda(auto:52)>; T = {}]::<lambda(auto:5&&)>; Result = future<>; T = {}]::<lambda(auto:2&&)> > /home/tgrabiec/src/scylla2/seastar/core/future.hh:690
0x16d0474 in then_wrapped<future<T>::handle_exception(Func&&) [with Func = sstables::sstable::~sstable()::<lambda(auto:52)>; T = {}]::<lambda(auto:5&&)>, future<> > /home/tgrabiec/src/scylla2/seastar/core/future.hh:880
0x1696e9c in handle_exception<sstables::sstable::~sstable()::<lambda(auto:52)> > /home/tgrabiec/src/scylla2/seastar/core/future.hh:1012
0x1638ba8 in sstables::sstable::~sstable() sstables/sstables.cc:1619
The leak is about allocations related to close() syscall tasks invoked
from sstable destructor, which were not waited for.
Message-Id: <1452783887-25244-1-git-send-email-tgrabiec@scylladb.com>
This reverts commit f0d68e4 ("main: start the http server in the first
step"). The service layer is not ready to serve clients before it's
fully up and running which causes early startup crashes everywhere.
Message-Id: <1452768015-22763-1-git-send-email-penberg@scylladb.com>
If authentication is disabled, nobody calls login() to set the current
user. There's untranslated code in client_state constructor to do just
that.
Fixes "You have not logged in" errors when USE statement is executed
with authentication disabled.
Message-Id: <1452759946-13998-1-git-send-email-penberg@scylladb.com>
"Add implementation of cassandra password authenticator, and user
password checking to CQL connections.
User/pwd are stored in system_auth table. Passwords are hashed
using glibc 'crypt_r'.
The latter is worth noting, as this is a difference compared to origin;
Origin uses Java bcrypt library for salt/hash, i.e. blowfish hashing.
Most glibc variants do _not_ have support for blowfish. To be 100%
compatible with imported origin tables we might need to add
bcrypt/blowfish sources into scylla (no packaged libs available afaict)
The code currently first attempts to use blowfish, if we happen to run
centos or Openwall, which has it compiled in. Otherwise we will fall
back to sha512, sha256 or even md5 depending on lib support.
To use:
* scylla.conf: authenticator=PasswordAuthenticator
* cqlsh -u cassandra -p cassandra
Not implemented (yet):
* "Authorizer", thus no KS/CF access checking
* CQL create/alter/delete user (create_user_statement etc). I.e. there is
only a single user name; default "cassandra:cassandra" user/pwd combo"
It's needed to keep the iterators valid in case eviciton is triggered
somehwere in between. It probably isn't because destructors should not
allocate, but better be safe.
Currently for wrap around the "begin" iterator would not meet with the
"end" iterator, invoking undefined behavior in erase_and_dispose()
which results in a crash.
Fixes#785
User db storage + login/pwd db using system tables.
Authenticator object is a global shard-shared singleton, assumed
to be completely immutable, thus safe.
Actual login authentication is done via locally created stateful object
(sasl challenge), that queries db.
Uses "crypt_r" for password hashing, vs. origins use of bcrypt.
Main reason is that bcrypt does not exist as any consistent package
that can be consumed, so to guarantee full compatibility we'd have
to include the source. Not hard, but at least initially more work than
worth.
Fixes#614
* Use warning threshold from config
* Don't throw exceptions. We're only supposed to warn.
* Try to actually estimate mutation data payload size, not
number of mutations.
Message-Id: <1452615759-23213-1-git-send-email-calle@scylladb.com>
Each segment chunk should contain column mappings for all schema
versions used by the mutations it contains. In order to avoid
duplication db::commitlog::segment remembers all schema versions already
written in current chunk.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Current commitlog interface requires writers to specify the size of a
new entry which cannot depend on the segment to which the entry is
written.
If column mappings are going to be stored in the commitlog that's not
enough since we don't know whether column mapping needs to be written
until we known in which segment the entry is going to be stored.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Refs #752
Paged aggregate queries will re-use the partition_slice object,
thus when setting a specific ck range for "last pk", we will hit
an exception case.
Allow removing entries (actually only the one), and overwriting
(using schema equality for keys), so we maintain the interface
while allowing the pager code to re-set the ck range for previous
page pass.
[tgrabiec: commit log cleanup, fixed issue ref]
Message-Id: <1452616259-23751-1-git-send-email-calle@scylladb.com>
Fixes#614
* Use warning threshold from config
* Don't throw exceptions. We're only supposed to warn.
* Try to actually estimate mutation data payload size, not
number of mutations.
Fixes#752
We set row limit for query to be min of page size/remaining in limit,
but if we have a multinode query we might end up with more rows than asked
for, so must do this again in post-processing.
Refs #792
Paged aggregate queries will re-use the partition_slice object,
thus when setting a specific ck range for "last pk", we will hit
an exception case.
Allow removing entries (actually only the one), and overwriting
(using schema equality for keys), so we maintain the interface
while allowing the pager code to re-set the ck range for previous
page pass.
v2:
* Changed to schema-equality checks so we sort of maintain a
sane api and behaviour, even with the 1-entry map
v3:
* Renamed remove "contains" in specific_ranges, and made the calling
code use more map-like logic, again to keep things cleaner
Fixes#752
We set row limit for query to be min of page size/remaining in limit,
but if we have a multinode query we might end up with more rows than asked
for, so must do this again in post-processing.
Message-Id: <1452606935-12899-2-git-send-email-calle@scylladb.com>
According to specification
(here https://wiki.apache.org/cassandra/InternodeEncryption)
when the internode encryption is set to `dc` the data passed between
DCs should be encrypted and similarly, when it's set to `rack`
the inter-rack traffic should encrypted.
Currently Scylla would encrypt the traffic inside a local DC in the
first case and inside the local RACK in the later one.
This patch fixes the encryption logic to follow the specification
above.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Message-Id: <1452501794-23232-1-git-send-email-vladz@cloudius-systems.com>
With Docker we might be running on a filesystem that does not support DMA
(aufs; or tmpfs on boot2docker), so let --developer-mode allow running
on those file systems.
Message-Id: <1452593083-25601-1-git-send-email-avi@scylladb.com>
yum command think "development-xxxx.xxxx" is older than "0.x", so nightly package mistakenly update with release version.
To prevent this problem, we should add greater number prior to "development".
Also same on Ubuntu package.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Message-Id: <1452592877-29721-1-git-send-email-syuu@scylladb.com>
Use systemd Type=notify to tell systemd about startup progress.
We can now use 'systemctl status scylla-server' to see where we are
in service startup, and 'systemctl start scylla-server' will wait until
either startup is complete, or we fail to start up.
read_entry did not verify that current chunk has enough data left
for a minimal entry. Thus we could try to read an entry from the slack
left in a chunk, and get lost in the file (pos > next, skip very much
-> eof). And also give false errors about corruption.
Message-Id: <1452517700-599-1-git-send-email-calle@scylladb.com>
This will add support for an user to clean up an entire keyspace
or some of its column families.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Cleanup is a procedure that will discard irrelevant keys from
all sstables of a column family, thus saving disk space.
Scylla will clean up a sstable by using compaction code, in
which this sstable will be the only input used.
Compaction manager was changed to become aware of cleanup, such
that it will be able to schedule cleanup requests and also know
how to handle them properly.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
The implementation is about storing generation of compacting sstables
in an unordered set per column family, so before strategy is called,
compaction manager will filter out compacting sstables.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Currently, compaction strategy is the responsible for both getting the
sstables selected for compaction and running compaction.
Moving the code that runs compaction from strategy to manager is a big
improvement, which will also make possible for the compaction manager
to keep track of which sstables are being compacted at a moment.
This change will also be needed for cleanup and concurrent compaction
on the same column family.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Cleanup is about rewriting a sstable discarding any keys that
are irrelevant, i.e. keys that don't belong to current node.
Parameter cleanup was added to compact_sstables.
If set to true, irrelevant code such as the one that updates
compaction history will be skipped. Logic was also added to
discard irrelevant keys.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
That code will be used by column family cleanup, so let's put
that code into a function. This change also improves the code
readability.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
"Our domain objects have schema version dependent format, for efficiency
reasons. The data structures which map between columns and values rely on
column ids, which are consecutive integers. For example, we store cells in a
vector where index into the vector is an implicit column id identifying table
column of the cell. When columns are added or removed the column ids may
shift. So, to access mutations or query results one needs to know the version
of the schema corresponding to it.
In case of query results, the schema version to which it conforms will always
be the version which was used to construct the query request. So there's no
change in the way query result consumers operate to handle schema changes. The
interfaces for querying needed to be extended to accept schema version and do
the conversions if necessary.
Shard-local interfaces work with a full definition of schema version,
represented by the schema type (usually passed as schema_ptr). Schema versions
are identified across shards and nodes with a UUID (table_schema_version
type). We maintain schema version registry (schema_registry) to avoid fetching
definitions we already know about. When we get a request using unknown schema,
we need to fetch the definition from the source, which must know it, to obtain
a shard-local schema_ptr for it.
Because mutation representation is schema version dependent, mutations of
different versions don't necessarily commute. When a column is dropped from
schema, the dropped column is no longer representable in the new schema. It is
generally fine to not hold data for dropped columns, the intent behind
dropping a column is to lose the data in that column. However, when merging an
incoming mutation with an existing mutation both of which have different
schema versions, we'd have to choose which schema should be considered
"latest" in order not to loose data. Schema changes can be made concurrently
in the cluster and initiated on different nodes so there is not always a
single notion of latest schema. However, schema changes are commutative and by
merging changes nodes eventually agree on the version. For example adding
column A (version X) on one node and adding column B (version Y) on another
eventually results in a schema version with both A and B (version Z). We
cannot tell which version among X and Y is newer, but we can tell that version
Z is newer than both X and Y. So the solution to the problem of merging
conflicting mutations could be to ensure that such merge is performed using
the schema which is superior to schemas of both mutations.
The approach taken in the series for ensuring this is as follows. When a node
receives a mutation of an unknown schema version it first performs a schema
merge with the source of that mutation. Schema merge makes sure that current
node's version is superior to the schema of incoming mutation. Once the
version is synced with, it is remembered as such and won't be synced with on
later mutations. Because of this bookkeeping, schema versions must be
monotonic; we don't want table altering to result in any earlier version
because that would cause nodes to avoid syncing with them. The version is a
cryptographically-secure hash of schema mutations, which should fulfill this
purpose in practice.
TODO: It's possible that the node is already performing a sync triggered by
broadcasted schema mutations. To avoid triggering a second sync needlessly, the
schema merging should mark incoming versions as being synced with.
Each table shard keeps track of its current schema version, which is
considered to be superior to all versions which are going to be applied to it.
All data sources for given column family within a shard have the same notion
of current schema version. Individual entries in cache and memtables may be at
earlier versions but this is hidden behind the interface. The entries are
upgraded to current version lazily on access. Sstables are immutable, so they
don't need to track current version. Like any other data source, they can be
queried with any schema version.
Note, the series triggered a bug in demangler:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68700"
* seastar d0bf6f8...ad3577b (9):
> httpd: close connection before deleting it
> reactor: support for non-O_DIRECT capable filesystems
> tests: modernize linecount
> IO queues: destruct within reactor's destructor
> tests: Use dnsdomainname in mkcert.gmk
> tests: memcached: workaround a possible race between flush_all and read
> apps: memcached: reduce the error during the expiration time translation
> timer: add missing #include
> core: do not call open_file_dma directly
Fixes#757.
The MOTD banner now printed upon .bash_profile execution,
if scylla is running, ends with a 'tput sgr0'. That command
appends an extra '[m' at the beginning of the output of any
following command. The automation scripts don't like this.
So let's add an 'echo' at the end of that path to add a newline,
avoiding the condition described above, and another one at the
'ScyllaDB is not started' path, for symmetry. I'm doing this
as it seems easier than having to develop heuristics to know
whether to remove or not that character.
CC: Shlomi Livne <slivne@scylladb.com>
Signed-off-by: Lucas Meneghel Rodrigues <lmr@scylladb.com>
Message-Id: <1452216044-28374-1-git-send-email-lmr@scylladb.com>
read_entry did not verify that current chunk has enough data left
for a minimal entry. Thus we could try to read an entry from the slack
left in a chunk, and get lost in the file (pos > next, skip very much
-> eof). And also give false errors about corruption.
Exceptions originated by an unimplemented to_string() methods
may interrupt the query() flow if not intercepted. Don't let it
happen.
Fixes issue #768
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
We want the statements to be removed before we ack the schema change,
otherwise it will race with all future operations.
Since the subscriber will be invoked on each shard, there is no need
to broadcast to all shards, we can just handle current shard.
Replicates https://issues.apache.org/jira/browse/CASSANDRA-7910 :
"Prepare a statement with a wildcard in the select clause.
2. Alter the table - add a column
3. execute the prepared statement
Expected result - get all the columns including the new column
Actual result - get the columns except the new column"
Currently the notify_*() method family broadcasts to all shards, so
schema merging code invokes them only on shard 0, to avoid doubling
notifications. We can simplify this by making the notify_*() methods
per-instance and thus shard-local.
If the schema_builder is constructed from an existing schema we need to
make sure that the original column ids of regular and static columns are
*not* used since they may become invalid if columns are added or
removed.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
When a column is dropped its name and deletion timestamp are added
to schema::_raw._dropped_columns to prevent data resurrection in case a
column with the same name is added. To reduce the number of lookups in
_dropped_columns this patch makes each instance of column_definition
to caches this information (i.e. timestamp of the latest removal of a
column with the same name).
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Knowing which columns were dropped (and when) is important to prevent
the data from the dropped ones reappearing if a new column is added with
the same name.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
We want the node's schema version to change whenever
table_schema_version of any table changes. The latter is calculated by
hashing mutations so we should also use mutation hash when calculating
schema digest.
We need to track which schema version were synced with on current node
to avoid triggering the sync on every mutation. We need to sync before
mutating to be able to apply the incoming mutation using current
node's schema, possibly applying irreverdible transformations to it to
make it conform.
There is one current schema for given column_family. Entries in
memtables and cache can be at any of the previous schemas, but they're
always upgraded to current schema on access.
The verb belongs to a seaprate client to avoid potential deadlocks
should the throttling on connection level be introduced in the
future. Another reason is to reduce latency for version requests as it
can potentially block many requests.
The intent is to make data returned by queries always conform to a
single schema version, which is requested by the client. For CQL
queries, for example, we want to use the same schema which was used to
compile the query. The other node expects to receive data conforming
to the requested schema.
Interface on shard level accepts schema_ptr, across nodes we use
table_schema_version UUID. To transfer schema_ptr across shards, we
use global_schema_ptr.
Because schema is identified with UUID across nodes, requestors must
be prepared for being queried for the definition of the schema. They
must hold a live schema_ptr around the request. This guarantees that
schema_registry will always know about the requested version. This is
not an issue because for queries the requestor needs to hold on to the
schema anyway to be able to interpret the results. But care must be
taken to always use the same schema version for making the request and
parsing the results.
Schema requesting across nodes is currently stubbed (throws runtime
exception).
Schema is tracked in memtable and cache per-entry. Entries are
upgraded lazily on access. Incoming mutations are upgraded to table's
current schema on given shard.
Mutating nodes need to keep schema_ptr alive in case schema version is
requested by target node.
We must use canonical_mutation form to allow for changes in the schema
of schema tables. The node which deserializes schema mutations may not
have the same version of the schema tables so we cannot use
frozen_mutation, which is a schema dependent form.
frozen_schema will transfer schema definition across nodes with schema
mutations. Because different nodes may have different versions of
schema tables, we cannot use frozen_mutations to transfer these
because frozen_mutation can only be read using the same version of the
schema it was frozen with. To solve this problem, new from of mutation
is introduced called canonical_mutation, which can be read using any
version of the schema.
This patch fixes a regression introduced by
a commit ca935bf "tests: Fix gossip_test".
database service initializes a replication_strategy
object and a replication_strategy requires a snitch
service to be initialized.
A snitch service requires a broadcast address to be
set.
If any of the above is not initialized we are going
to hit the corresponding assert().
Set a snitch to a SimpleSnitch and a broadcast
address to 127.0.0.1.
Fixes issue #770
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Message-Id: <1452421748-9605-1-git-send-email-vladz@cloudius-systems.com>
The goal is to provide various test cases with a way of iterating over
many combinations of mutaitons. It's good to have this in one place to
avoid duplication and increased coverage.
Right now in some places we use column_id, and in some places
size_t. Solve it by using column_count_type whose meaning is "an
integer sufficiently large for indexing columns". Note that we cannot
use column_id because it has more meaning to it than that.
The version needs to change value not only on structural changes but
also temporal. This is needed for nodes to detect if the version they
see was already synchronized with or not even if it has the same
structure as the past versions. We also need to end up with the same
version on all nodes when schema changes are commuted.
For regular mutable schemas version will be calculated from underlying
mutations when schema is announced. For static schemas of system
keyspace it is calculated by hashing scylla version and column id,
because we don't have mutations at the time of building the schema.
We will be able to reuse the code in frozen_schema. We need to read
data in mutation form so that we can construct the correct
schema_table_version, and attach the mutations to schema_ptr.
It simplifies add_table_to_schema_mutation() interface.
The current code is also a bit confusing, partition_key is created
with the keyspaces() schema and used in mutations destined for the
columnfamilies() schema. It works, the types are the same, but looks a
bit scary.
For static and regular (row) columns it is very convenient in some
cases to utilize the fact that columns ordered by ids are also ordered
by name. It currently holds, so make schema export this guarantee and
enable consumers to rely on.
The static schema::row_column_ids_are_ordered_by_name field is about
allowing code external to schema to make it very explicit (via
static_assert) that it relies on this guarantee, and be easily
discoverable in case we would have to relax this.
With 10 sstables/shard and 50 shards, we get ~10*50*50 messages = 25,000
log messages about sstables being ignored. This is not reasonable.
Reduce the log level to debug, and move the message to database.cc,
because at its original location, the containing function has nothing to
do with the message itself.
Reviewed-by: Raphael S. Carvalho <raphaelsc@cloudius-systems.com>
Message-Id: <1452181687-7665-1-git-send-email-avi@scylladb.com>
Wait for the future returned by the http server start process to resolve,
so we know it is started. If it doesn't, we'll hit the or_terminate()
further down the line and exit with an error code.
Message-Id: <1452092806-11508-3-git-send-email-avi@scylladb.com>
Because our shutdown process is crippled (refs #293), we won't shutdown the
snitch correctly, and the sharded<> instance can assert during shutdown.
This interferes with the next patch, which adds orderly shutdown if the http
server fails to start.
Leak it intentionally to work around the problem.
Message-Id: <1452092806-11508-2-git-send-email-avi@scylladb.com>
Make the following tests pass:
bootstrap_test.py:TestBootstrap.shutdown_wiped_node_cannot_join_test
bootstrap_test.py:TestBootstrap.killed_wiped_node_cannot_join_test
1) start node2
2) wait for cql connection with node2 is ready
3) stop node2
4) delete data and commitlog directory for node2
5) start node2
In step 5), node2 will do the bootstrap process since its data,
including the system table is wiped. It will think itself is a completly
new node and can possiblly stream from wrong node and violate
consistency.
To fix, we reject the boot if we found the node was in SHUTDOWN or
STATUS_NORMAL.
CASSANDRA-9765
Message-Id: <47bc23f4ce1487a60c5b4fbe5bfe9514337480a8.1452158975.git.asias@scylladb.com>
Implement the wait for gossip to settle logic in the bootup process.
CASSANDRA-4288
Fixes:
bootstrap_test.py:TestBootstrap.shutdown_wiped_node_cannot_join_test
1) start node2
2) wait for cql connection with node2 is ready
3) stop node2
4) delete data and commitlog directory for node2
5) start node2
In step 5, sometimes I saw in shadow round of node2, it gets node2's
status as BOOT from other nodes in the cluster instead of NORMAL. The
problem is we do not wait for gossip to settle before we start cql server,
as a result, when we stop node2 in step 3), other nodes in the cluster
have not got node2's status update to NORMAL.
The previous SSL enablement patches do make uses of these
options but they are still marked as Unused.
Change this and also update the db/config.hh documentation
accordingly.
Syntax is now:
client_encryption_options:
enabled: true
certificate: <path-to-PEM-x509-cert> (default conf/scylla.crt)
keyfile: <path-to-PEM-x509-key> (default conf/scylla.key)
Fixes: #756.
Signed-off-by: Benoît Canet <benoit@scylladb.com>
Message-Id: <1452032073-6933-1-git-send-email-benoit@scylladb.com>
Compaction fixes from Raphael:
There were two problems causing issue 676:
1) max_purgeable was being miscalculated (fixed by b7d36af).
2) empty row not being removed by mutation_partition::do_compact
Testcase is added to make sure that a tombstone will be purged under
certain conditions.
do_compact() wasn't removing an empty row that is covered by a
tombstone. As a result, an empty partition could be written to a
sstable. To solve this problem, let's make trim_rows remove a
row that is considered to be empty. A row is empty if it has no
tombstone, no marker and no cells.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
"This is another version of the repair overhaul, to avoid streaming *all* the
data between nodes by sending checksums of token ranges and only streaming
ranges which contain differing data."
Support the "hosts" and "dataCenters" parameters of repair. The first
specifies the known good hosts to repair this host from (plus this host),
and the second asks to restrict the repair to the local data center (you
must issue the repair to a node in the data center you want to repair -
issuing the command to a data center other than the named one returns
an error).
For example these options are used by nodetool commands like:
nodetool repair -hosts 127.0.0.1,127.0.0.2 keyspace
nodetool repair -dc datacenter1
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
The existing repair code always streamed the entire content of the
database. In this overhaul, we send "repair_checksum_range" messages to
the other nodes to verify whether they have exactly the same data as
this node, and if they do, we avoid streaming the identical code.
We make an attempt to split the token ranges up to contain an estimated
100 keys each, and send these ranges' checksums. Future versions of this
code will need to improve this estimation (and make this "100" a parameter)
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
This patch adds a function sync_range() for synchronizing all partitions
in a given token range between a set of replicas (this node and a list of
neighbors).
Repair will call this function once it has decided that the data the
replicas hold in this range is not identical.
The implementation streams all the data in the given range, from each of
the neighbors to this node - so now this node contains the most up-to-date
data. It then streams the resulting data back to all the neighbors.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
This patch adds a new type of message, "REPAIR_CHECKSUM_RANGE" to scylla's
"messaging_service" RPC mechanism, for the use of repair:
With this message the repair's master host tells a slave host to calculate
the checksum of a column-family's partitions in a given token range, and
return that checksum.
The implementation of this message uses the checksum_range() function
defined in the previous patch.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
This patch adds functions for calculating the checksum of all the
partitions in a given token range in the given column-family - either
in the current shard, or across all shards in this node.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
This patch adds a mechanism for calculating a checksum for a set of
partitions. The repair process will use these checksums to compare the
data held by different replicas.
We use a strong checksum (SHA-256) for each individual partition in the set,
and then a simple XOR of those checksums to produce a checksum for the
entire set. XOR is good enough for merging strong checksums, and allows us
to independently calculate the checksums of different subsets of the
original sets - e.g., each shard can calculate its own checksum and we
can XOR the resulting checksums to get the final checksum.
Apache Cassandra uses a very similar checksum scheme, also using SHA-256
and XOR. One small difference in the implementation is that we include the
partition key in its checksum, while Cassandra don't, which I believe to
have no real justification (although it is very unlikely to cause problems
in practice). See further discussion on this in
https://issues.apache.org/jira/browse/CASSANDRA-10728.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
A cut-and-paste accident in query::to_partition_range caused the wrong
end's inclusiveness to be tested.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Everything except alter_table_statement::announce_migration() is
translated. announce_migration() has to wait for multi schema support to
be merged.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
We have an API that wraps open_file_dma which we use in some places, but in
many other places we call the reactor version directly.
This patch changes the latter to match the former. It will have the added benefit
of allowing us to make easier changes to these interfaces if needed.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
Message-Id: <29296e4ec6f5e84361992028fe3f27adc569f139.1451950408.git.glauber@scylladb.com>
This exception was not caught properly as a std::exception
by report_failed_future call to report_exception because the
superclass std::exception was not initialized.
Fixes#669.
Signed-off-by: Benoît Canet <benoit@scylladb.com>
Recently, Scylla was changed to mandate the use of XFS
for its data directories, unless the flag --developer-mode true
is provided. So during the AMI setup stage, if the user
did not provide extra disks for the setup scripts to prepare,
the scylla service will refuse to start. Therefore, the
message in scylla_prepare has to be changed to an actual
error message, and the file name, to be changed to
something that reflects the event that happened.
Signed-off-by: Lucas Meneghel Rodrigues <lmr@scylladb.com>
Avi Kivity (1):
Merge "rpc negotiation fixes" from Gleb
Gleb Natapov (3):
rpc: fix peer address printing during logging
rpc: make server send negotiation frame back before closing connection on error
rpc: fix documentation for negotiation procedure.
Nadav Har'El (1):
fix operator<<() for std::vector<T>
Tomasz Grabiec (1):
core/byteorder: Add missing include
scylla-ami.sh moved some ami specific files. This parts have been
dropped when converging scylla-ami into scylla_install. Fixing that.
Signed-off-by: Shlomi Livne <shlomi@scylladb.com>
The script imports the /etc/sysconfig/scylla-server for configuration
settings (NR_PAGES). The /etc/sysconfig/scylla-server iincludes an AMI
param which is of string value and called as a last step in
scylla_install (after scylla_bootparam_setup has been initated).
The AMI variable is setup in scylla_install and is used in multiple
scripts. To resolve the conflict moving the import of
/etc/sysconfig/scylla-server after the AMI variable has been compared.
Fixes: #744
Signed-off-by: Shlomi Livne <shlomi@scylladb.com>
Switch to CentOS 7 as the Docker base image. It's more stable and
updated less frequently than Fedora. As a bonus, it's Thrift package
doesn't pull the world as a dependency which reduces image size from 700
MB to 380 MB.
Suggested by Avi.
Message-Id: <1451911969-26647-1-git-send-email-penberg@scylladb.com>
If requests are delayed downstream from the cql server, and the client is
able to generate unrelated requests without limit, then the transient memory
consumed by the requests will overflow the shard's capacity.
Fix by adding a semaphore to cap the amount of transient memory occupied by
requests.
Fixes#674.
compatible: can be cast, keeps sort order
value-compatible: can be cast, may change sort order
frozen: values participate in sort order
unfrozen: only sort keys participate in sort order
Fixes#740.
"Messaging service inherits from seastar::async_sharded_service which
guaranties that sharded<>::stop() will not complete until all references
to the service will not go away. It was done specifically to avoid using
more verbose gate interface in sharded<> services since it turned out
that almost all of them need one eventually. Unfortunately patches to
add redundant gate to messaging_service sneaked pass my review. The series
reverts them."
When 'nodetool clearsnapshot' is given no parameters it should
remove all existing snapshots.
Fixes issue #639
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
service::storage_service::clear_snapshot() was built around _db.local()
calls so it makes more sense to move its code into the 'database' class
instead of calling _db.local().bla_bla() all the time.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Current service initialization is a total mess in cql_test_env. Start
the service the same order as in main.cc.
Fixes#715, #716
'./test.py --mode release' passes.
It's just a waste of time to find them manually
when compiling ScyllaDB on a fresh install: add them.
Also fix the ninja-build name.
Signed-off-by: Benoît Canet <benoit@scylladb.com>
After upgrading an AMI and trying to stop and start a machine the
/var/lib/scylla/coredump is not created. Create the directory if it does
not exist prior to generating core
Signed-off-by: Shlomi Livne <shlomi@scylladb.com>
There's two untranslated grammar subrules in the 'relation' rule. We
have all the necessary AST classes translated, so translate the
remaining subrules.
Refs #534.
Signed-off-by: Pekka Enberg <penberg@scylladb.com>
messaging_service will use private ip address automatically to connect a
peer node if possible. There is no need for the upper level like
streaming to worry about it. Drop it simplifies things a bit.
The stream info was creted but was left out of the stream state. This
patch adds the created stream_info to the stream state vector.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
* seastar 8b2171e...de112e2 (4):
> build: disable -fsanitize=vptr harder
> Merge "Make RPC more robust against protocol changes" from Gleb
> new when_all implementation
> apps: memcached: Don't fiddle with the expiration time value in a usable range
Commit 2ba4910 ("main: verify that the NOFILE rlimit is sufficient")
added a recommendation to set NOFILE rlimit to 200k. Update our release
binaries to do the same.
"This series solve an issue with the load broadcaster that reports negative
values due to an integer wrap around. While fixing this issue an additional
change was made so that the load_map would return doubles and not formatted
string. This is a better API, safer and better documented."
Since commit 16596385ee, long_token() is already checking
t.is_minimum(), so the comment which explains why it does not (for
performance) is no longer relevant. And we no longer need to check
t._kind before calling long_token (the check we do here is the same
as is_minimum).
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Check the list of column families passed as an option to repair, to
provide the user with a more meaningful exception when a non-existant
column family is passed.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
This was a plain bug - ranges_opt is supposed to parse the option into
the vector "var", but took the vector by value.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Support the "columnFamilies" parameter of repair, allowing to repair
only some of the column families of a keyspace, instead of all of them.
For example, using a command like "nodetool repaire keyspace cf1 cf2".
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
The script scylla_coredump_setup was introduced in
9b4d0592, and added to the scylla rpm spec file, as a
post script. However, calling yum when there's one
yum instance installng scylla server will cause a deadlock,
since yum waits for the yum lock to be released, and the
original yum process waits for the script to end.
So let's remove this from the script. Debian shouldn't be
affected, since it was never added to the debian build
rules (to the best of my knowlege, after analyzing 9b4d0592),
hence I did not remove it. It should cause the same problem
with apt-get in case it was used.
CC: Takuya ASADA <syuu@scylladb.com>
[ penberg: Rebase and drop statement about 'abrt' package not in Fedora. ]
Signed-off-by: Lucas Meneghel Rodrigues <lmr@scylladb.com>
A default value was not set for the "incremental" and "parallelism"
repair parameters, so Scylla can wrongly decide that they have an
unsupported value.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
The repair API use to have an undocumented parameter list similiar to
origin.
This patch changes the way repair is getting its parameters.
Instead of a one undocumented string it now lists all the different
optional parameters in the swagger file and accept them explicitely.
Reviewed-by: Nadav Har'El <nyh@scylladb.com>
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
This change set the http server to start as the first step in the boot
order.
It is helpfull if some other step takes a long time or stuck.
Fixes#725
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Actually check that a snapshot directory with a given tag
exists instead of just checking that a 'snapshot' directory
exists.
Fixes issue #689
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
In origin the storage_serivce report the load map as a formatted string.
As an API a better option is to report the load map as double and let
the JMX proxy do the formatting.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
The map_reduce0 convert the result value to the init value type. In
load_bradcaster 0 is of type int.
This result with an int wrap around and negative results.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Add partial support for the "incremental" option (only support the
"false" setting, i.e., not incremental repair) and the "parallelism"
option (the choice of sequential or parallel repair is ignored - we
always use our own technique).
This is needed because scylla-jmx passes these options by default
(e.g., "incremental=false" is passed to say this is *not* incremental
repair, and we just need to allow this and ignore it).
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
When throwing an "unsupported repair options" exception to the caller
(such as "nodetool repair"), also list which options were not recognized.
Additionally, list the options when logging the repair operation.
This patch includes an operator<< implementation for pretty-printing an
std::unordered_map. We may want to move it later to a more central
location - even Seastar (like we have a pretty-printer for std::vector
in core/sstring.hh).
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
max_purgeable was being incorrectly calculated because the code
that creates vector of uncompacted sstables was wrong.
This value is used to determine whether or not a tombstone can
be purged.
Operand < is supposed to be used instead in the callback passed
as third parameter to boost::set_difference.
This fix is a step towards closing the issue #676.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
The start_native_transport() function in storage_service expects the
'enabled' option to be defined. If the option is not defined, it means
that encryption is implicitly disabled.
Fixes#718.
"Adds support for TLS/SSL encrypted (and cert verified)
connections for message service
* Modify config option to match "native" style cerificate management
* Add SSL options to messaging service and generate SSL server/client
endpoints when required
* Add config option handling to init/main"
* Massage user options in main
* Use them in storage_service, and if needed, load certificates etc
and pass to transport/cql server.
Conflicts:
service/storage_service.cc
Optional credentials argument determine if SSL or normal
server socket is created.
Note: This does not follow the pattern of "socket as argument", simply
because this is a distributed object, so only trivial or immutable
objects should be passed to it.
Describe scylla version of option.
Note, for test usage, the below should be workable:
server_encryption_options:
internode_encryption: all
certificate: seastar/tests/test.crt
truststore: seastar/tests/catest.pem
keyfile: seastar/tests/test.key
Since the seastar test suite contains a snakeoil cert + trust
combo
* Accept port + credentials + option for what to encrypt
* If set, enable a SSL listener at ssl_port
* Check outgoing connections by IP to determine if
they should go to SSL/normal endpoint
Requires seastar RPC patch
Note: currently, the connections created by messaging service
does _not_ do certificate name verification. While DNS lookup
is probably not that expensive here, I am not 100% sure it is
the desired behaviour.
Normal trust is however verified.
* Mark option used
* Make sub-options adapted to seastar-tls useable values (i.e. x509)
Syntax is now:
server_encryption_options:
internode_encryption: <none, all, dc, rack>
certificate: <path-to-PEM-x509-cert> (default conf/scylla.crt)
keyfile: <path-to-PEM-x509-key> (default conf/scylla.key)
truststore: <path-to-PEM-trust-store-file> (default empty,
use system trust)
"When a node gain or regain responsibility for certain token ranges, streaming
will be performed, upon receiving of the stream data, the row cache
is invalidated for that range.
Refs #484."
The describe_ring method in storage_service did not report the start and
end tokens.
Also for rpc addresses that are not the local address, it returned the
value representation (including the version) and not just the adress.
Fixes#695
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Use steady_clock instead of high_resolution_clock where monotonic
clock is required. high_resolution_clock is essentially a
system_clock (Wall Clock) therefore may not to be assumed monotonic
since Wall Clock may move backwards due to time/date adjustments.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Use steady_clock instead of high_resolution_clock where monotonic
clock is required. high_resolution_clock is essentially a
system_clock (Wall Clock) therefore may not to be assumed monotonic
since Wall Clock may move backwards due to time/date adjustments.
Fixes issue #638
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
"Fixes stream_session hangs:
1) if the sending node is gone, the receiving peer will wait forever
2) if the node which should send COMPLETE_MESSAGE to the peer node is gone,
the peer node will wait forever"
This patch fixes a bug where the *first* run of "nodetool repair" always
returned immediately, instead of waiting for the repair to complete.
Repair operations are asynchronous: Starting a repair returns a numeric
id, which can then be used to query for the repair's completion, and this
is what "nodetool repair" does (through our JMX layer). We started with
the repair ID "0", the next one is "1", and so on.
The problem is that "nodetool repair", when it sees 0 being returned,
treats it not as a regular repair ID, but rather as an answer that
there is nothing to repair - printing a message to that effect and *not*
waiting for the repair (which was correctly started) to complete.
The trivial fix is to start our repair IDs at 1, instead of 0.
We currently do not return 0 in any case (we don't know there is nothing
to repair before we actually start the work, and parameter errors
cause an exception, not a return of 0).
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
If we can't open the file, we will fail with a misterious error. It is a costumary
scenario, though, since people who are unaware or have just forgotten about seastar's
restriction of direct io access may put those files in tmpfs and other mount points.
We have a direct_io check that is designed exactly for this purpose, so as to give
the user a better error message. This patch makes use of it.
Fixes#644
Signed-off-by: Glauber Costa <glauber@scylladb.com>
It is hard-coded as 30 seconds at the moment.
Usage:
$ scylla --ring-delay-ms 5000
Time a node waits to hear from other nodes before joining the ring in
milliseconds.
Same as -Dcassandra.ring_delay_ms in cassandra.
The midpoint() algorithm to find a token between two tokens doesn't
work correctly in case of wraparound. The code tried to handle this
case, but did it wrong. So this patch fixes the midpoint() algorithm,
and adds clearer comments about why the fixed algorithm is correct.
This patch also modifies two midpoint() tests in partitioner_test,
which were incorrect - they verified that midpoint() returns some expected
values, but expected values were wrong!
We also add to the test a more fundemental test of midpoint() correctness,
which doesn't check the midpoint against a known value (which is easy to
get wrong, like indeed happened); Rather we simply check that the midpoint
is really inside the range (according to the token ordering operator).
This simple test failed with the old implementation of midpoint() and
passes with the new one.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
The problem is that we set the session state to WAIT_COMPLETE in
send_complete_message's continuation, the peer node might send
COMPLETE_MESSAGE before we run the continuation, thus we set the wrong
status in COMPLETE_MESSAGE's handler and will not close the session.
Before:
GOT STREAM_MUTATION_DONE
receive task_completed
SEND COMPLETE_MESSAGE to 127.0.0.2:0
GOT COMPLETE_MESSAGE, from=127.0.0.2, connecting=127.0.0.3, dst_cpu_id=0
complete: PREPARING -> WAIT_COMPLETE
GOT COMPLETE_MESSAGE Reply
maybe_completed: WAIT_COMPLETE -> WAIT_COMPLETE
After:
GOT STREAM_MUTATION_DONE
receive task_completed
maybe_completed: PREPARING -> WAIT_COMPLETE
SEND COMPLETE_MESSAGE to 127.0.0.2:0
GOT COMPLETE_MESSAGE, from=127.0.0.2, connecting=127.0.0.3, dst_cpu_id=0
complete: WAIT_COMPLETE -> COMPLETE
Session with 127.0.0.2 is complete
If the session is idle for 10 minutes, close the session. This can
detect the following hangs:
1) if the sending node is gone, the receiving peer will wait forever
2) if the node which should send COMPLETE_MESSAGE to the peer node is
gone, the peer node will wait forever
Fixes simple_kill_streaming_node_while_bootstrapping_test.
Get from address from cinfo. It is needed to figure out which stream
session this mutation is belonged to, since we need to update the keep
alive timer for this stream session.
Currently, if the node is actually down, although the streaming_timeout
is 10 seconds, the sending of the verb will return rpc_closed error
immediately, so we give up in 20 * 5 = 100 seconds. After this change,
we give up in 10 * 30 = 300 seconds at least, and 10 * (30 + 30) = 600
seconds at most.
It is oneway message at the moment. If a COMPLETE_MESSAGE is lost, no
one will close the session. The first step to fix the issue is to try to
retransmit the message.
$NAME is full name of distribution, for script it is too long.
$ID is shortened one, which is more useful.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
The helper function for summing statistic over the column family are
template function that infer the return type acording to the type of the
Init param.
In the API the return value should be int64_t, passing an integer would
cause a number wrap around.
A partial output from the nodetool cfstats after the fix
nodetool cfstats keyspace1
Keyspace: keyspace1
Read Count: 0
Read Latency: NaN ms.
Write Count: 4050000
Write Latency: 0.009178098765432099 ms.
Pending Flushes: 0
Table: standard1
SSTable count: 12
Space used (live): 1118617445
Space used (total): 23336562465
Fixes#682
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
From Calle:
Fixes#589
Query should not return dangling static row in partition without any
regular/ck columns if a CK restriction is applied.
Refs #650
Fixes bug in CK range code for paging, and removes CK use for tables with not
clustering -> way simpler code. Also removed lots of workaround code no longer
required.
Note that this patch set does not fully fix #650/paging since bug #663 causes
duplicate rows. Still almost there though.
Boost::date_time doesn't accept some of the date and time formats that
the origin do (e.g. 2013-9-22 or 2013-009-22).
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Refs #640
* Remove use of cluster key range for tables without CK
Checking CK existance once and use the info allows us to remove some
stupid complexity in checking for "last key" match
* With fix for #589 we can also remove some superfluous code to
compensate for that issue, and make "partition end" simper
* Remove extra row in CK case. Not needed anymore
End result is that pager now more or less only relies on adapted query
ranges.
timestamp_from_string() is used by both timestamp and date types, so it
is better to move the try { } catch { } to the functions itself instead
of expecting its callers to catch exceptions.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
On AMI, scylla-server fails to systemctl restart because scylla_prepare tries to mount /var/lib/scylla even it's already mounted.
This patch fixes the issue.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
* seastar b44d729...51154f7 (6):
> semaphore: add with_semaphore()
> scripts: posix_net_conf.sh: don't transform wide CPU mask
> resource: fix build for systems without HWLOC
> build: link libasan before all other libraries
> Use sys_membarrier() when available
> build: add missing library (boost_filesystem)
Fixes#589
If we got no rows, but have live static columns, we should only
give them back IFF we did not have any CK restrictions.
If ck:s exist, and we have a restriction on them, we either have maching
rows, or return nothing, since cql does not allow "is null".
This patch introduces a test for reading keys from a single sstable with
the range begining and end being the keys present in the index summary.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
When choosing a relevant range of buckets it wasn't taken into account
whether the range bounds are inclusive or not. That may have resulted in
more buckets being read than necessary which was a condition not
expected by the code responsible from looking for a relevant keys inside
the buckets.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
If a sstable doesn't belong to current shard, mark_for_deletion
should be called for the deletion manager to still work.
It doesn't mean that the sstable will be deleted, but that the
sstable is not relevant to the current shard, thus it can be
deleted by the deletion manager in the future.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
When a node gain or regain responsibility for certain token ranges,
streaming will be performed, upon receiving of the stream data, the
row cache is invalidated for that range.
Refs #484.
From Paweł:
"This series fixes sstables::key_reader not respecting range inclusiveness
if the bounds were the keys that were present in the index summary.
Fixes #663."
This patch introduces a test for reading keys from a single sstable with
the range begining and end being the keys present in the index summary.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
When choosing a relevant range of buckets it wasn't taken into account
whether the range bounds are inclusive or not. That may have resulted in
more buckets being read than necessary which was a condition not
expected by the code responsible from looking for a relevant keys inside
the buckets.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
"This series adds support for nodetool command 'drain'. The general idea
of this command is to close all connection (both with clienst and other
nodes) and flush all memtables to disk.
Fixes #662."
"Merge AMI scripts to dist/common/scripts, make it usable on non-AMI
environments. Provides a script to do all settings automatically, which
able to run as one-liner like this:
curl http://url_to_scylla_install | sudo bash -s -- -d /dev/xvdb,/dev/xvdc -n eth0 -l ./
Also enables coredump, save it to /var/lib/scylla/coredump"
When the server is shutting down a flag _stopping is set and listeners
are aborted using abort_accept(), which causes accept() calls to return
failed futures. However, accept handler just checks that the flag
_stopping is set and returns which causes a failed future to be
destroyed and a warning is printed.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
The underlying data source for cache should not be the same memtable
which is later used to update the cache from. This fixes the following
assertion failure:
row_cache_test_g: utils/logalloc.hh:289: decltype(auto) logalloc::allocating_section::operator()(logalloc::region&, Func&&) [with Func = memtable::make_reader(schema_ptr, const partition_range&)::<lambda()>]: Assertion `r.reclaiming_enabled()' failed.
The problem is that when memtable is merged into cache their regions
are also merged, so locking cache's region locks the memtable region
as well.
Currently sourcing for the second time causes an exception from
pretty printer registration:
Traceback (most recent call last):
File "./scylla-gdb.py", line 41, in <module>
gdb.printing.register_pretty_printer(gdb.current_objfile(), build_pretty_printer())
File "/usr/share/gdb/python/gdb/printing.py", line 152, in register_pretty_printer
printer.name)
RuntimeError: pretty-printer already registered: scylla
Fixes the case where background activity needed to complete CL=ONE writes
is queued up in the storage proxy, and the client adds new work faster
than it can be cleared.
The last two loops were incorrectly inside the first one. That's a
bug because a new sstable may be emplaced more than once in the
sstable list, which can cause several problems. mark_for_deletion
may also be called more than once for compacted sstables, however,
it is idempotent.
Found this issue while auditing the code.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
"* seastar 294ea30...b44d729 (5):
> Merge "Properly distribute IO queues" from Glauber
> reactor: allow more poll time in virtualized environments
> reactor: fix idle-poll limit
> reactor: use a vector of unique_ptr for the IO queues
> io queues: make the queues really part of the reactor"
With consistency level less then ALL mutation processing can move to
background (meaning client was answered, but there is still work to
do on behalf of the request). If background request rate completion
is lower than incoming request rate background request will accumulate
and eventually will exhaust all memory resources. This patch's aim is
to prevent this situation by monitoring how much memory all current
background request take and when some threshold is passed stop moving
request to background (by not replying to a client until either memory
consumptions moves below the threshold or request is fully completed).
There are two main point where each background mutation consumes memory:
holding frozen mutation until operation is complete in order to hint it
if it does not) and on rpc queue to each replica where it sits until it's
sent out on the wire. The patch accounts for both of those separately
and limits the former to be 10% of total memory and the later to be 6M.
Why 6M? The best answer I can give is why not :) But on a more serious
note the number should be small enough so that all the data can be
sent out in a reasonable amount of time and one shard is not capable to
achieve even close to a full bandwidth, so empirical evidence shows 6M
to be a good number.
I am sure it's a compiler issue but I am not ready to give up and
upgrade just yet:
sstables/compaction.cc:307:55: error: converting to ‘std::unordered_map<int, long int>’ from initializer list would use explicit constructor ‘std::unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::unordered_map(std::unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::size_type, const hasher&, const key_equal&, const allocator_type&) [with _Key = int; _Tp = long int; _Hash = std::hash<int>; _Pred = std::equal_to<int>; _Alloc = std::allocator<std::pair<const int, long int> >; std::unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::size_type = long unsigned int; std::unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::hasher = std::hash<int>; std::unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::key_equal = std::equal_to<int>; std::unordered_map<_Key, _Tp, _Hash, _Pred, _Alloc>::allocator_type = std::allocator<std::pair<const int, long int> >]’
stats->start_size, stats->end_size, {});
Test was failing because _qp (distributed<cql3::query_processor>) was stopped
before _db (distributed<database>).
Compaction manager is member of database, and when database is stopped,
compaction manager is also stopped. After a2fb0ec9a, compaction updates the
system table compaction history, and that requires a working query context.
We cannot simply move _qp->stop() to after _db->stop() because the former
relies on migration_manager and storage_proxy. So the most obvious fix is to
clean the global variable that stores query context after _qp was stopped.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
The previous patch added message_service read()/write() support for all
types which know how to serialize themselves through our "old" serialization
API (serialize()/deserialize()/serialized_size()).
So we no longer need the almost 200 lines of repetitive code in
messaging_service.{cc,hh} which defined these read/write templates
separately for a dozen different types using their *serialize() methods.
We also no longer need the helper functions read_gms()/write_gms(), which
are basically the same code as that in the template functions added in the
previous patch.
Compilation is not significantly slowed down by this patch, because it
merely replaces a dozen templates by one template that covers them all -
it does not add new template complexity, and these templates are anyway
instantiated only in messaging_service.cc (other code only calls specific
functions defined in messaging_service.cc, and does not use these templates).
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Currently, messaging_service only supports sending types for which a read/
write function has been explicitly implemented in messageing_service.hh/cc.
Some types already have serialization/deserialization methods inside them,
and those could have been used for the serialization without having to write
new functions for each of these types. Many of these types were already
supported explicitly in messaging_service.{cc,hh}, but some were forgot -
for example, dht::token.
So this patch adds a default implemention of messaging_service write()/read()
which will work for any type which has these serialization methods.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
* seastar 5b9e3da...294ea30 (9):
> Merge "IO queues" from Glauber
> reactor: increment check_direct_io_support to also deal with files
> Merge "SSL/TLS initial certificate validation" from Calle
> tutorial.md: remove inaccurate statements about x86
> build: verify that the installed compiler is up to date
> build: complain if fossil version of gnutls is installed
> build: fix debian naming of gnutls-devel package
> build: add configure-time check for gnutls-devel
> tutorial.md: introduction to asynchrnous programming
send_to_live_endpoints() is never waited upon, it does its job in the
background. This patch formalize that by changing return value to void
and also refactoring code so that frozen_mutation shared pointer is not
held more that it should: currently it is held until send_mutation()
completes, but since send_mutation() does not use frozen_mutation
asynchronously this is not necessary.
Replace db_clock::now_in_usec() and db_clock::now() * 1000 accesses
where the intent is to create a new auto-generate cell timestamp with
a call to new_timestamp(). Now the knowledge of how to create timestamps
is in a single place.
"get_compactions returns progress information for each compaction
running in the system. It can be accessed using swagger UI.
'nodetool compactionstats' is not working yet because of some
pending work in the nodetool side."
Apparently, link hook copy constructor is a no-op and move contructor
doesn't exist so the code is correct, but that explicit move makes code
needlessly confusing.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
That's important for compaction stats API that will need stats
data of each ongoing compaction.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
This list will store compaction_stats for each ongoing compaction.
That's why register and deregister methods are provided.
This change is important for compaction stats API that needs data
of each ongoing compaction, such as progress, ks, cf, etc.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
"This patchset will make Scylla update the system table
COMPACTION_HISTORY whenever a compaction job finishes.
Functions were added to both update and retrieve the
content of this system table. Compaction history API
is also enabled in this series."
When compaction job finishes, call function to update the system
table COMPACTION_HISTORY. That's also needed for the compaction
history API.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
This method is intended to return content of the system table
COMPACTION_HISTORY as a vector of compaction_history_entry.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
If a sstable doesn't belong to current shard, mark_for_deletion
should be called for the deletion manager to still work.
It doesn't mean that the sstable will be deleted, but that the
sstable is not relevant to the current shard, thus it can be
deleted by the deletion manager in the future.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
There is a check whose intent was to detect wrap around during walk of
the ring tokens by comparing the split point with minimum token, which
is supposed to be inserted by the ring iterator. It assumed that when
we encounter it, the range is a wrap around. It doesn't hold when
minimum token is part of the token metadata or set of tokens is empty.
In such case, a full range would be split into 3 overlapping full
ranges. The fix is to drop the assumption and instead ensure that
ranges do not wrap around by unwrapping them if necessary.
Fixes#655.
The default move assignment operator calls boost::intrusive::set's move
assignment operator, which leaks, because it does not believe it owns
the data.
Fix by providing a custom implementation.
All components of prefixable compound type are preceeded by their
length what makes them not byte order comparable.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Frozen collection type names must be wrapped in FrozenType so that we
are able to store the types correctly in system tables.
This fixes#646 and fixes#580.
Signed-off-by: Pekka Enberg <penberg@scylladb.com>
The test_assignement() function is invoked via the Cassandra unit tests
so we might as well implement it.
Signed-off-by: Pekka Enberg <penberg@scylladb.com>
It is 30 seconds instead of 5 seconds by default. To align with c*.
Pleas note, after this a node will takes at least 30 seconds to complete
a bootstrap.
Originally, large allocation test case attempted to allocate an object
as big as halft of the space used by the lsa. That failed when the test
was executed with lower amount of memory available mainly due to the
memory fragmentation caused by previous test cases.
This patches reduces the size of the large allocation to 3/8 of the
total space used by the lsa which is still a lot but seems to make the
test pass even with as little memory as 64MB per shard.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
If we get a core dump from a user, it is important to be able to
identify its version. Copy the release string into the heap (which is
copied into the code dump), so we can search for it using the "strings"
or "ident" commands.
Reviewed-by: Nadav Har'El <nyh@scylladb.com>
This fixes compile error:
In function `logalloc::segment_zone::segment_zone()':
/home/lmr/Code/scylla/utils/logalloc.cc:412: undefined reference to `logalloc::segment_zone::minimum_size'
collect2: error: ld returned 1 exit status
ninja: build stopped: subcommand failed.
Signed-off-by: Lucas Meneghel Rodrigues <lmr@scylladb.com>
blob_storage defined with attribute packed which makes its alignment
requirement equal 1. This means that its members may be unaligned.
GCC is obviously aware of that and will generate appropriate code
(and not generate ubsan checks). However, there are few places where
members of blob_storage are accessed via pointers, these have to be
wrapped by unaligned_cast<> to let the compiler know that the location
pointed to may be not aligned properly.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Form Paweł:
This series fixes support for clustering keys which trailing components
are null. The solution is to use clustering_key_prefix instead of
clustering_key everywhere.
Fixes#515.
Schemas using compact storage can have clustering keys with the trailing
components not set and effectively being a clustering key prefixes
instead of full clustering keys.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
In case of non-compound dense tables the column name is just the value
of the clustering key (which has only one component). Current code just
casts clustering_key to bytes_view which works because there is no
additional metadata in single element clustering keys.
However, that may change when the internal representation of clustering
key is changed so explicitly extract the proper component.
This change will become necessary when clustering_key is replaced by
clustering_key_prefix.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
In case of schemas that use compact storage it is possible that trailing
components of clustering keys are not set.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
When this tool was written, we were still using /var/lib/cassandra as a default
location. We should update it.
Signed-off-by: Glauber Costa <glauber@scylladb.com>
* seastar 5dc22fa...c5e595b (3):
> memory: be less strict about NUMA bindings
> reactor: let the resource code specify the default memory reserve
> resource: reserve even more memory when hwloc is compiled in
Fixes#642
"This series attempts to make LSA more friendly for large (i.e. bigger
than LSA segment) allocations. It is achieved by introducing segment
zones – large, contiguous areas of segments and using them to allocate
segments instead of calling malloc() directly.
Zones can be shrunk when needed to reclaim memory and segments can be
migrated either to reduce number of zone or to defragment one in order
to be able to shrink it. LSA tries to keep all segments at the lower
addresses and reclaims memory starting from the zones in the highest
parts of the address space."
Also:
[PATCH scylla v1 0/7] gossip mark node down fix + cleanup
[PATCH scylla v1 0/2] Refuse decommissioned node to rejoin
[PATCH scylla] storage_service: Fix added node not showing up in nodetool in status joining
When replacing a node, we might ignore the tokens so that the tokens is
empty. In this case, we will have
std::unordered_map<inet_address, std::unordered_set<token>> = {ip, {}}
passed to token_metadata::update_normal_tokens(std::unordered_map<inet_address,
std::unordered_set<token>>& endpoint_tokens)
and hit the assert
assert(!tokens.empty());
1) Start node 1, node 2, node 3
2) Stop node 3
3) Start node 4 to replace node 3
4) Kill node 4 (removal of node 3 in system.peers is not flushed to disk)
5) Start node 4 (will load node 3's token and host_id info in bootup)
This makes
"Token .* changing ownership from 127.0.0.3 to 127.0.0.4"
messages printed again in step 5) which are not expected, which fails the dtest
FAIL: replace_first_boot_test (replace_address_test.TestReplaceAddress)
----------------------------------------------------------------------
Traceback (most recent call last):
File "scylla-dtest/replace_address_test.py",
line 220, in replace_first_boot_test
self.assertEqual(len(movedTokensList), numNodes)
AssertionError: 512 != 256
In commit 56df32ba56 (gossip: Mark node as
dead even if already left). A node liveness check is missed.
Fix it up.
Before: (mark a node down multiple times)
[Tue Dec 8 12:16:33 2015] INFO [shard 0] gossip - InetAddress 127.0.0.3 is now DOWN
[Tue Dec 8 12:16:33 2015] DEBUG [shard 0] storage_service - endpoint=127.0.0.3 on_dead
[Tue Dec 8 12:16:34 2015] INFO [shard 0] gossip - InetAddress 127.0.0.3 is now DOWN
[Tue Dec 8 12:16:34 2015] DEBUG [shard 0] storage_service - endpoint=127.0.0.3 on_dead
[Tue Dec 8 12:16:35 2015] INFO [shard 0] gossip - InetAddress 127.0.0.3 is now DOWN
[Tue Dec 8 12:16:35 2015] DEBUG [shard 0] storage_service - endpoint=127.0.0.3 on_dead
[Tue Dec 8 12:16:36 2015] INFO [shard 0] gossip - InetAddress 127.0.0.3 is now DOWN
[Tue Dec 8 12:16:36 2015] DEBUG [shard 0] storage_service - endpoint=127.0.0.3 on_dead
After: (mark a node down only one time)
[Tue Dec 8 12:28:36 2015] INFO [shard 0] gossip - InetAddress 127.0.0.3 is now DOWN
[Tue Dec 8 12:28:36 2015] DEBUG [shard 0] storage_service - endpoint=127.0.0.3 on_dead
The only reason we needed it is to make
_application_state[key] = value
work.
With the current default constructor, we increase the version number
needlessly. To fix and to be safe, remove the default constructor
completely.
Backport: CASSANDRA-8801
a53a6ce Decommissioned nodes will not rejoin the cluster.
Tested with:
topology_test.py:TestTopology.decommissioned_node_cant_rejoin_test
The get_token_endpoint API should return a map of tokens to endpoints,
including the bootstrapping ones.
Use get_local_storage_service().get_token_to_endpoint_map() for it.
$ nodetool -p 7100 status
Status=Up/Down
|/ State=Normal/Leaving/Joining/Moving
-- Address Load Tokens Owns Host ID Rack
UN 127.0.0.1 12645 256 ? eac5b6cf-5fda-4447-8104-a7bf3b773aba rack1
UN 127.0.0.2 12635 256 ? 2ad1b7df-c8ad-4cbc-b1f1-059121d2f0c7 rack1
UN 127.0.0.3 12624 256 ? 61f82ea7-637d-4083-acc9-567e0c01b490 rack1
UJ 127.0.0.4 ? 256 ? ced2725e-a5a4-4ac3-86de-e1c66cecfb8d rack1
Fixes#617
Originally, lsa allocated each segment independently what could result
in high memory fragmentation. As a result many compaction and eviction
passes may be needed to release a sufficiently big contiguous memory
block.
These problems are solved by introduction of segment zones, contiguous
groups of segments. All segments are allocated from zones and the
algorithm tries to keep the number of zones to a minimum. Moreover,
segments can be migrated between zones or inside a zone in order to deal
with fragmentation inside zone.
Segment zones can be shrunk but cannot grow. Segment pool keeps a tree
containing all zones ordered by their base addresses. This tree is used
only by the memory reclamer. There is also a list of zones that have
at least one free segments that is used during allocation.
Segment allocation doesn't have any preferences which segment (and zone)
to choose. Each zone contains a free list of unused segments. If there
are no zones with free segments a new one is created.
Segment reclamation migrates segments from the zones higher in memory
to the ones at lower addresses. The remaining zones are shrunk until the
requested number of segments is reclaimed.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
A dynamic bitset implementation that provides functions to search for
both set and cleared bits in both directions.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Currently test case "Testing reading when memory can't be reclaimed."
assumes that the allocation section used by row cache upon entering
will require more free memory than there is available (inc. evictable).
However, the reserves used by allocation section are adjusted
dynamically and depend solely on previous events. In other words there
is no guarantee that the reserve would be increased so much that the
allocation will fail.
The problem is solved by adding another allocation that is guaranteed
to be bigger than all evictable and free memory.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Scattering of blobs from Avi:
This patchset converts the stack to scatter managed_bytes in lsa memory,
allowing large blobs (and collections) to be stored in memtable and cache.
Outside memtable/cache, they are still stored sequentially, but it is assumed
that the number of transient objects is bounded.
The approach taken here is to scatter managed_bytes data in multiple
blob_storage objects, but to linearize them back when accessing (for
example, to merge cells). This allows simple access through the normal
bytes_view. It causes an extra two copies, but copying a megabyte twice
is cheap compared to accessing a megabyte's worth of small cells, so
per-byte throughput is increased.
Testing show that lsa large object space is kept at zero, but throughput
is bad because Scylla easily overwhelms the disk with large blobs; we'll
need Glauber's throttling patches or a really fast disk to see good
throughput with this.
Add linearize() and unlinearize() methods that allow making an
atomic_cell_or_collection object temporarily contiguous, so we can examine
it as a bytes_view.
Instead of allocating a single blob_storage, chain multiple blob_storage
objects in a list, each limited not to exceed the allocation_strategy's
max_preferred_allocation_size. This allows lsa to allocate each blob_storage
object as an lsa managed object that can be migrated in memory.
Also provide linearize()/scatter() methods that can be used to temporarily
consolidate the storage into a single blob_storage. This makes the data
contiguous, so we can use a regular bytes_view to examine it.
This adds the implementation for the index_summary_off_heap_memory for a
single column family and for all of them.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Similiar to origin, off heap memory, memory_footprint is the size of
queus multiply by the structure size.
memory_footprint is used by the API to report the memory that is taken
by the summary.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
If there is no snapshot directory for the specific column family,
get_snapshot_details should return an empty map.
This patch check that a directory exists before trying to iterate over
it.
Fixes#619
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Scylla changes:
sstable.cc: Remove file_exists() function which conflicts with seastar's
Amnon Heiman (2):
reactor: Add file_exists method
Add a wrapper for file_exists
Avi Kivity (2):
Merge "Introduce shared_future" from Tomasz
Merge ""scripts: a few fixes in posix_net_conf.sh" from Vlad
Gleb Natapov (3):
rpc: not stop client in error state
avoid allocation in parallel_for_each is there is nothing to do
memory: fix size_to_idx calculation
Nadav Har'El (1):
test: fix use-after-free in timertest
Pawe�� Dziepak (1):
memory: use size instead of old_size to shrink memory block
Tomasz Grabiec (7):
file: Mark move constructor as noexcept
core: future: Add static asserts about type's noexcept guarantees
core: future: Drop now redundant move_noexcept flag
core: future_state: Make state getters non-destructive for non-rvalue-refs
core: future: Make get_available_state() noexcept
core: Introduce shared_future
Make json_return_type movable
Vlad Zolotarov (8):
scripts: posix_net_conf.sh: ban NIC IRQs from being moved by irqbalance
scripts: posix_net_conf.sh: exclude CPU0 siblings from RPS
scripts: posix_net_conf.sh: Configure XPS
scripts: posix_net_conf.sh: Add a new mode for MQ NICs
scripts: posix_net_conf.sh: increase some backlog sizes
core: to_sstring(): cleanup
core: to_sstring_strintf(): always use %g(or %lg) format for floating point values
core: prevent explicit calls for to_sstring_sprintf()
In a recent discussion with the XFS developers, Dave Chinner recommended
us *not* to use discard, but rather issue fstrims explicitly. In machines
like Amazon's c3-class, the situation is made worse by the fact that discard
is not supported by the disk. Contrary to my intuition, adding the discard
mount option in such situation is *not* a nop and will just create load
for no reason.
Signed-off-by: Glauber Costa <glommer@scylladb.com>
Objects extending json_base are not movable, so we won't be able to
pass them via future<>, which will assert that types are nothrow move
constructible.
This problem only affects httpd::utils_json::histogram, which is used
in map-reduce. This patch changes the aggregation to work on domain
value (utils::ihistrogram) instead of json objects.
Our premier allocation_strategy, lsa, prefers to limit allocations below
a tenth of the segment size so they can be moved around; larger allocations
are pinned and can cause memory fragmentation.
Provide an API so that objects can query for this preferred size limit.
For now, lsa is not updated to expose its own limit; this will be done
after the full stack is updated to make use of the limit, or intermediate
steps will not work correctly.
The config file expresses this number in MB, while total_memory() gives us
a quantity in bytes. This causes the commitlog not to flush until we reach
really skyhigh numbers.
While we need this fix for the short term before we cook another release,
I will note that for the mid/long term, it would be really helpful to stop
representing memory amounts as integers, and use an explicit C++ type for
those. That would have prevented this bug.
Signed-off-by: Glauber Costa <glommer@scylladb.com>
Print a map in the form of [(]{ key0 : value0 }[, { keyN : valueN }]*[)]
The map is printed inside () brackets if it's frozen.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
In origin, there are two APIs to get the information about the current
running compactions. Both APIs do the string formatting.
This patch changes the API to have a single API get_compaction that
would return a list of summary object.
The jmx would do the string formatting for the two APIs.
This change gives a better API experience is it's better documented and
would make it easier to support future format changes in origin.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
That's what we're trying to standardize on.
This patch also fixes an issue with current query::result::serialize()
not being const-qualified, because it modifies the
buffer. messaging_service did a const cast to work this around, which
is not safe.
This patch adds the implementation to the get_version.
After this patch the following url will be available:
messaging_service/version?addr=127.0.0.1
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
"This series allows the compaction manager to be used by the nodetool as a stub implementation.
It has two changes:
* Add to the compaction manager API a method that returns a compaction info
object
* Stub all the compaction method so that it will create an unimplemented
warning but will not fail, the API implementation will be reverted when the
work on compaction will be completed."
This patch fixes the following cql_query_test failure.
cql_query_test: scylla/seastar/core/sharded.hh:439:
Service& seastar::sharded<Service>::local() [with Service =
gms::gossiper]: Assertion `local_is_initialized()' failed.
The problem is in gossiper::stop() we call gossip::add_local_application_state()
which will in turn call gms::get_local_gossiper(). In seastar::sharded::stop
_instances[engine().cpu_id()].service = nullptr;
return inst->stop().then([this, inst] {
return _instances[engine().cpu_id()].freed.get_future();
});
We set the _instances to nullptr before we call the stop method, so
local_is_initialized asserts when we try to access get_local_gossiper
again.
To fix, we make the stopping of gossiper explicit. In the shutdown
procedure, we call stop_gossiping() explicitly.
This has two more advantages:
1) The api to stop gossip is now calling the stop_gossiping() instead of
sharing the seastar::sharded's stop method.
2) We can now get rid of the _handler seastar::sharded helper.
The add interface of the estimated histogram is confusing as it is not
clear what units are used.
This patch removes the general add method and replace it with a add_nano
that adds nanoseconds or add that gets duration.
To be compatible with origin, nanoseconds vales are translated to
microseconds.
This patch adds a started counter, that is used to mark the number of
operation that were started.
This counter serves two purposes, it is a better indication for when to
sample the data and it is used to indicate how many pending operations
are.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
This patch adds the column family API that return the snapshot size.
The changes in the swagger definition file follo origin so the same API will be used for the metric and the
column_family.
The implementation is based on the get_snapshot_details in the
column_family.
This fix:
425
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Backport: CASSANDRA-10330
ae4cd69 Print versions for gossip states in gossipinfo
For instance, the version for each state, which can be useful for
diagnosing the reason for any missing states. Also instead of just
omitting the TOKENS state, let's indicate whether the state was actually
present or not.
With
Node 1 (Seed node, Port 7000 is opened, 10.184.9.144)
Node 2 (Port 7000 is opened, 10.184.9.145)
Node 3 (Port 7000 is blocked by firewall)
On Node 3, we saw the following error which was very confusing: Node 3
saw Node 1 and Node 3 but it complained it can not contact any seeds.
The message "Node 10.184.9.144 is now part of the cluster" and friends
are actually messages printed during the gossip shadow round where Node
3 connects to Node 1's port 7000 and Node 1 returns all info it knows to
Node 3, so that Node 3 knows Node 1 and Node 2 and we see the "Node
10.184.9.144/145 is now part of the cluster" message.
However, during the normal gossip round, Node 3 will not mark Node 1 and
Node 2 UP until the Seed node initiates a gossip round to Node 3, (note
port 7000 on node 3 is blocked in this case). So Node 3 will not mark
Node 1 and Node 2 UP and we see the "Unable to contact any seeds" error.
[shard 0] storage_service - Loading persisted ring state
[shard 0] gossip - Node 10.184.9.144 is now part of the cluster
[shard 0] gossip - inet_address 10.184.9.144 is now UP
[shard 0] gossip - Node 10.184.9.145 is now part of the cluster
[shard 0] gossip - inet_address 10.184.9.145 is now UP
[shard 0] storage_service - Starting up server gossip
scylla_run[12479]: Start gossiper service ...
[shard 0] storage_service - JOINING: waiting for ring information
[shard 0] storage_service - JOINING: schema complete, ready to bootstrap
[shard 0] storage_service - JOINING: waiting for pending range calculation
[shard 0] storage_service - JOINING: calculation complete, ready to bootstrap
[shard 0] storage_service - JOINING: getting bootstrap token
[shard 0] storage_service - JOINING: sleeping 5000 ms for pending range setup
scylla_run[12479]: Exiting on unhandled exception of type 'std::runtime_error': Unable to contact any seeds!
Backported: CASSANDRA-8336 and CASSANDRA-9871
84b2846 remove redundant state
b2c62bb Add shutdown gossip state to prevent timeouts during rolling restarts
8f9ca07 Cannot replace token does not exist - DN node removed as Fat Client
Fixes:
When X is shutdown, X sends SHUTDOWN message to both Y and Z, but for
some reason, only Y receives the message and Z does not receive the
message. If Z has a higher gossip version for X than Y has for
X, Z will initiate a gossip with Y and Y will mark X alive again.
X ------> Y
\ /
\ /
Z
Fixes: #593
"Changes the parser/replayer to treat data corruption as non-fatal,
skipping as little as possible to get the most data out of a segment,
but keeping track of, and reporting, the amount corrupted.
Replayer handles this and reports any non-fatal errors on replay finish.
Also added tests for corruption cases.
This patch series contains a cleanup-patch for commitlog_tests that was
previously submitted, but got lost."
If something bad happens between write request handler creation and
request execution the request handler have to be destroyed. Currently
code tries to do that explicitly in all places where request may be
abandoned, but it misses some (at least one). This patch replaces this
by introducing unique_response_handler object that will remove the handler
automatically if request is not executed for some reason.
Rename antlr3-tool to antlr3 (same as distribution package), and use distribution version if it's available
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
"Before this change, populations could race with update from flushed
memtable, which might result in cache being populated with older
data. Populations started before the flush are not considering the
memtable nor its sstable.
The fix employed here is to make update wait for populations which
were started before the flushed memtable's sstable was added to the
undrelying data source. All populatinos started after that are
guaranteed to see the new data. The update() call will wait only for
current populating reads to complete, it will not wait for readers to
get advanced by the consumer for instance."
To avoid a race where natural endpoint was updated to contain node A,
but A was not yet removed from pending endpoints.
This fixes the root cause of commit d9d8f87c1 (storage_proxy: filter out
natural endpoints from pending endpoint). This patch alone fixes#539,
but we still want commit d9d8f87c1 to be safe.
When other bootstrapping/leaving/moving nodes are found during
bootstrap, instead of throwing immediately, sleep and try again for one
minute, hoping other nodes will finish the operation soon.
Since we are retrying using shadow gossip round more than once, we need
to put the gossip state back to shadow round after each shadow round, to
make shadow round works correctly.
This is useful when starting an empty cluster for testing. E.g,
$ scylla --listen-address 127.0.0.1
$ sleep 3
$ scylla --listen-address 127.0.0.2
$ sleep 3
$ scylla --listen-address 127.0.0.3
Without this patch, node 3 will hit the check.
TIME STATUS
-----------------------
Node 1:
32:00 Starts
32:00 In NORMAL status
Node 2:
32:03 Starts
32:04 In BOOT status
32:10 In NORMAL status
Node 3:
32:06 Starts
32:06 Found node 2 in BOOT status, hit the check, sleep and try again
32:11 Found node 2 in NORMAL status, can keep going now
32:12 In BOOT status
32:18 In NORMAL status
When other bootstrapping/leaving/moving nodes are found during
bootstrap, instead of throwing immediately, sleep and try again for one
minute, hoping other nodes will finish the operation soon.
This is useful when starting an empty cluster for testing. E.g,
$ scylla --listen-address 127.0.0.1
$ scylla --listen-address 127.0.0.2
$ scylla --listen-address 127.0.0.3
Without this patch, node 3 will hit the check.
TIME STATUS
-----------------------
Node 1:
25:19 Starts
25:20 In NORMAL status
Node 2:
25:19 Starts
25:23 In BOOT status
25:28 In NORMAL status
Node 3:
25:19 Starts
25:24 Found node 2 in BOOT status, hit the check, sleep and try again
25:29 Found node 2 in NORMAL status, can keep going now
25:29 In BOOT status
25:34 In NORMAL status
Before this change, populations could race with update from flushed
memtable, which might result in cache being populated with older
data. Populations started before the flush are not considering the
memtable nor its sstable.
The fix employed here is to make update wait for populations which
were started before the flushed memtable's sstable was added to the
undrelying data source. All populatinos started after that are
guaranteed to see the new data.
The text data type is no longer present in CQL binary protocol v3 and
later. We don't need it for encoding earlier versions either because
it's an alias for varchar which is present in all CQL binary protocol
versions.
Fixes#526.
Signed-off-by: Pekka Enberg <penberg@scylladb.com>
This patch plus pekka's previous commit 3c72ea9f96
"gms: Fix gossiper::handle_major_state_change() restart logic"
fix CASSANDRA-7816.
Backported from:
def4835 Add missing follow on fix for 7816 only applied to
cassandra-2.1 branch in 763130bdbde2f4cec2e8973bcd5203caf51cc89f
763130b Followup commit for 7816
2199a87 Fix duplicate up/down messages sent to native clients
Tested by:
pushed_notifications_test.py:TestPushedNotifications.restart_node_test
CQL 3.2.1 introduces a "TRUNCATE TABLE X" alias for "TRUNCATE X":
4e3555c1d9
Fix our CQL grammar to also support that.
Please note that we don't bump up advertised CQL version yet because our
cqlsh clients won't be able to connect by default until we upgrade them
to C* 2.1.10 or later.
Fixes#576
Signed-off-by: Pekka Enberg <penberg@scylladb.com>
The FIXMEs are no longer valid, we load schema on bootstrap and don't
support hot-plugging of column families via file system (nor does
Cassandra).
Handling of missing tables matches Cassandra 2.1, applies log
it and continue, queries propagate the error.
If request comes after natural endpoint was updated to contain node A,
but A was not yet removed from pending endpoints it will be in both and
write request logic cannot handle this properly. Filter nodes which are
already in natural endpoint from pending endpoint to fix this.
Fixes#539.
boost::heap::binomial_heap allocates helper object in push() and,
therefore, may throw an exception. This shouldn't happen during
compaction.
The solution is to reserve space for this helper object in
segment_descriptor and use a custom allocator with
boost::heap::binomial_heap.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
LSA memory reclaimer logic assumes that the amount of memory used by LSA
equals: segments_in_use * segment_size. However, LSA is also responsible
for eviction of large objects which do not affect the used segmentcount,
e.g. region with no used segments may still use a lot of memory for
large objects. The solution is to switch from measuring memory in used
segments to used bytes count that includes also large objects.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Since this won't check disk types, may re-initialize RAID on EBS when first block was lost.
But in such condition, probably re-initialize RAID is the only choice we can take, so this is fine.
Fixes#364.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
With this patch, start two nodes
node 1:
scylla --rpc-address 127.0.0.1 --broadcast-rpc-address 127.0.0.11
node 2:
scylla --rpc-address 127.0.0.2 --broadcast-rpc-address 127.0.0.12
On node 1:
cqlsh> SELECT rpc_address from system.peers;
rpc_address
-------------
127.0.0.12
which means client should use this address to connect node 2 for cql and
thrift protocol.
It is same as
-Dcassandra.consistent.rangemovement
in cassandra.
Use it as:
$ scylla --consistent-rangemovement 0
or
$ scylla --consistent-rangemovement 1
Messaging service closes connection in rpc call continuation on
closed_error, but the code runs for each outstanding rpc call on the
connection, so first continuation may destroy genuinely closed connection,
then connection is reopened and next continuation that handless previous
error kills now perfectly healthy connection. Fix this by closing
connection only in error state.
From Avi:
Origin supports a notion of empty values for non-container types; these
are serialized as zero-length blobs. They are mostly useless and only
retained for compatibility.
The implementation here introduces a wrapper maybe_empty<T>, similar to
optional<T> but oriented towards usually-nonempty usage with implicit
conversion.
There is more work needed for full empty support: fixing up deserializers to
create empty values instead of nulls, and splitting up data_value into
data_value and a data_value_nonnull for the cases that require it.
(I chose maybe_empty<> rather than using optional<data_value> for nullable
data_value both because it requires fewer changes, and because
optional<data_value> introduces a lot of control flow when moving or copying,
which would be mostly useless in most cases).
This cleanup patch got lost in git-space some time ago. It is however sorely
needed...
* Use cleaner wrapper for creating temp dir + commit log, avoiding
having to clear and clean in every test, etc.
* Remove assertions based on file system checks, since these are not
valid due to both the async nature of the CL, and more to the point,
because of pre-allocation of files and file blocks. Use CL
counters/methods instead
* Fix some race conditions to ensure tests are safe(r)
* Speed up some tests
Discern fatal and non-fatal excceptions, and handle data corruption
by adding to stats, resporting it, but continue processing.
Note that "invalid_arguement", i.e. attempting to replay origin/old
segments are still considered fatal, as it is probably better to
signal this strongly to user/admin
Parser object now attempts to skip past/terminate parsing on corrupted
entries/chunks (as detected by invalid sizes/crc:s). The amount of data
skipped is kept track of (as well as we can estimate - pre-allocation
makes it tricky), and at the end of parsing/reporting, IFF errors
occurred, and exception detailing the failures is thrown (since
subsciption has little mechanism to deal with this otherwise).
Thus a caller can decide how to deal with data corruption, but will be
given as many entries as possible.
An empty serialized representation means an empty value, not NULL.
Fix up the confusion by converting incorrect make_null() calls to a new
make_empty(), and removing make_null() in empty-capable types like
bytes_type.
Collections don't support empty serialized representations, so remove
the call there.
Paramter evaluation order is unspecified, so it's possible that the
move of 'schema' into lambda captures would happen before construction of
mutation.
"To speed up boot, parallelism was introduced to our code that loads
sstables from a column family, a function was implemented to read
the minimum from a sstable to determine whether it belongs to the
current shard, and buffer size in read simple is dynamically chosen
based on the size of the file and dma alignment.
The latter is important because filter file can be considerably
large when the respective sstable (data file) is very large.
Before this patchset, scylla took about 5 minutes to boot with a
data directory of 660GB. After this patchset, scylla took about 20
seconds to boot with the same data directory."
Avi says:
"A small buffer size will hurt if we read a large file, but
a large buffer size won't hurt if we read a small file, since
we close it immediately."
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Currently, we only determine if a sstable belongs to current shard
after loading some of its components into memory. For example,
filter may be considerably big and its content is irrelevant to
decide if a sstable should be included to a given shard.
Start using the functions previously introduced to optimize the
sstable loading process. add_sstable no longer checks if a sstable
is relevant to the current shard.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Boot may be slow because the function that loads sstables do so
serially instead of in parallel. In the callback supplied to
lister::scan_dir, let's push the future returned by probe_file
(function that loads sstable) into a vector of future and wait
for all of them at the end.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
We cannot share some dependency package names between 14.04 and 15.10, so need to add ifdefs.
Not tested on other version of Ubuntu.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Origin supports (https://issues.apache.org/jira/browse/CASSANDRA-5648) "empty"
values even for non-container types such as int. Use maybe_empty<> to
encapsulate abstract_type::native_type, adding an empty flag if needed.
Similar to optional<>, with the following differences:
- decays back to the encapsulated type, with an emptiness check;
this reflects the expectation that the value will rarely be empty
- avoids conditionals during copy/move (and requires a default constructor),
again with the same expectation.
When we start to sending mutations for cf_id to remote node, remote node
might do not have the cf_id anymore due to dropping of the cf for
instance.
We should not fail the streaming if this happens, since the cf does not
exist anymore there is no point streaming it.
Fixes#566
When a new node joins a cluster, it will starts a gossip round with seed
node. However, within this round, the seed node will not tell the new
node anything it knows about other nodes in the cluster, because the
digest in the gossip SYN message contains only the new node itself and
no other nodes. The seed node will pick randomly from the live nodes,
including the newly added node in do_gossip_to_live_member to start a
gossip round. If the new node is "lucky", seed node will talk to it very
soon and tells all the information it knows about the cluster, thus the
new node will mark the seed node alive and think it has seen the seed
node. If there considerably large number of live nodes, it might take a
long time before the seed node pick the new node and talk to it.
In bootstrap code, storage_service::bootstrap checks if we see any nodes
after sleep of RING_DELAY milliseconds and throw "Unable to contact any
seeds!" if not, thus the node will fail to bootstrap.
To help the seed node talk to new node faster, we favor new node in
do_gossip_to_live_member.
In origin, get_all_endpoint_states perform all the information
formatting and returns a string.
This is not a good API approach, this patch replaces the implementation
so the API will return an array of values and the JMX will do the
formatting.
This is a better API and would make it simpler in the future to stay in
sync with origin output.
This patch is part of #508
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Fixes#551.
Change mountpoint to /var/lib/scylla, copy conf/ on it.
Note: need to replace conf/ with symlink to /etc/scylla when new rpm uploaded on yum repository.
Signed-off-by: Takuya ASADA <syuu@scylladb.com>
Signed-off-by: Pekka Enberg <penberg@iki.fi>
If we get a partition with no row data, but statics, we should treat this as
a row (include in count), but also make sure we skip to next partition
if our page ends here.
The "end partition" with zero rows but static data can also happen if we
happen to resume paging by giving a column range exluding all data. In this
case we should _not_ include it, since we have already provided the
data in question in previous page.
Fixes#556
1.) Should not reset to input query state if run repeatedly
2.) And if run repeatedly without input state, likewise keep
the internal one active
Fixes#560
"To keep compatibility with scylla-tools-java, it links /etc/scylla to /var/lib/scylla/conf.
Problem on this patchset is, I added SCYLLA_HOME and SCYLLA_CONF on /etc/sysconfig/scylla-server.
However, the file is marked as config file, it won't be automatically upgrade.
If user doesn't upgrade the file manually, scylla-server still able to run with /var/lib/scylla/conf because we have symlink, but never switches to /etc/scylla."
While the objects above max_manage_object_size aren't stored in the
LSA segments they are still considered to be belonging to the LSA
region and are evictable using that region evictor.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
"This series adds the natural_endpoints API. It adds the implementation to the storage_service and to the storage_service API.
After this series the noodtool command getendpoints should work.
example:
$ bin/nodetool getendpoints keyspace1 standard1 0x5032394c323239385030127.0.0.2
127.0.0.2"
This patch adds the API for timeout messages and dropped messages.
For dropped messages, origin has two APIs one for messages and one for
command.
droped messages return the number of messages per ver, so our API was
rename to reflect that.
For dropped messages (command) we currently do not have this logic of
throwing messages before sending, so the API will always return 0.
The total timeout API was removed and will be done on the jmx proxy
level.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
If listen_address is different than broadcast_address, we should use
broadcast_address for the seeds list. Check and ask user to fix the
configuration, e.g.,
$ scylla --rpc-address 127.0.0.1 --listen-address 127.0.0.1 --broadcast-address 192.168.1.100 --seed-provider-parameters seeds=127.0.0.1
Use broadcast_address instead of listen_address for seeds list: seeds={127.0.0.1}, listen_address=127.0.0.1, broadcast_address=192.168.1.100
Exiting on unhandled exception of type 'std::runtime_error': Use broadcast_address for seeds list
Write handler keeps track of all endpoints that not yet acked mutation
verb. It uses broadcast address as an enpoint id, but if local address
is different from broadcast address for local enpoints acknowledgements
will come from different address, so socket address cannot be used as
an acknowledgement source. Origin solves this by sending "from" in each
message, it looks like an overhead, solve this by providing endpoint's
broadcast address in rpc client_info and use that instead.
The restart logic is wrong because C* had a bug in
bf599fb5b062cbcc652da78b7d699e7a01b949ad and they fixed later and we
translated the broken version. We must check if there is an existing
endpoint state and call on_restart() hooks on that, not the newly
available endpoint state.
Spotted while inspecting the code.
Acked-by: Asias He <asias@scylladb.com>
From Avi:
Memtables do not use an allocating_section to guard against allocation
failure, and hence can fail an allocation. Reproducible by changing
perf_mutation to use an allocating type (bytes_type with a nontrivial
size) and making the loop longer.
Fix by using an allocating_section.
Recently, I have introduced cf_stats into the database, propagating all the way
back to the column family. The problem, however, is that some tests create a
column family config themselves instead of going through make_column_family.
That is ultimately ok if those tests are not expected to flush memtables. But
if they are, the cf_stats pointer will be null and we will crash. Although
there are many solutions to this, the one that is in tune with our current
practices is to have the test that requires it provide an empty cf_stats storage
area that can be written to. That's already how we handle the disk directory and
other things like compaction properties.
With this patch, test.py passes again.
Signed-off-by: Glauber Costa <glommer@scylladb.com>
This patch substitutes uint64_t for uint32_t as the type for
commitlog_total_space_in_mb. Moving to 64 is not strictly needed, since even a
signed 32-bit type would allow us to easily handle 2TB. But since we store that
in the commitlog as a 64-bit value, let's match it.
Moving from unsigned to signed, however, allow us to represent negative
numbers. With that in place, we can change the semantics of the value
slightly, so to allow a negative number to mean "all memory".
The reason behind this, is that the default value "8GB", is an artifact of the
JVM. We don't need that, and in many-shards configuration, each shard flushes
the commitlog way too often, since 8GB / many_shards = small_number.
8GB also happens to be a popular heap size for C* in the JVM. For us, we would
like to equate that (at least) with the amount of memory. The problem is how to
do that without introducing new options or changing the semantics of existing
options too radically.
The proposed solution will allow us to still parse C* yaml files, since those
will always have positive numbers, while introducing our own defaults.
Signed-off-by: Glauber Costa <glommer@scylladb.com>
Debian package system has two types of package, 'native' and 'non-native'.
'native' is the package just for Debian, it contains debian/ directory source tar.gz, doesn't have debian.tar.gz.
'non-native' has orig.tar.gz which is upstream source code tar ball, then it has debian.tar.gz which contains debian/ directory.
Scylla is 'native' now but should be 'non-native' since this is not just for Debian, so move debian/ to dist/ubuntu/, make orig.tar.gz using git-archive-all, copy dist/ubuntu/debian/ to debian/ then generate debian.tar.gz.
atomic_cell will soon become type-aware, so add helpers to class operation
that can supply the type, as it is available in operation::column.type.
(the type will be used in following patches)
schema_tables manages some boolean columns stored in system tables; it
dynamically creates them from C++ values. But as we lacked bool->data_value
conversion, the C++ value was converted to a int32_type. Somehow this didn't
cause any problems, but with some pending patches I have, it does.
Add a bool->data_value converting constructor to fix this.
Since bytes is a very generic value that is returned from many calls,
it is easy to pass it by mistake to a function expecting a data_value,
and to get a wrong result. It is impossible for the data_value constructor
to know if the argument is a genuine bytes variable, a data_value of another
type, but serialized, or some other serialized data type.
To prevent misuse, make the data_value(bytes) constructor
(and complementary data_value(optional<bytes>) explicit.
When do_stop_native_transport exits, cserver is destroyed which can
happen before cserver->stop(). Fix by capturing cserver in
cserver->stop()'s continuation to extend its lifetime. The same for
thrift server.
scylla: scylla/seastar/core/sharded.hh:327: seastar::sharded<Service>::~sharded()
[with Service = transport::cql_server]: Assertion `_instances.empty()' failed.
When analyzing a recent performance issue, I found helpful to keep track of
the amount of memtables that are currently in flight, as well as how much memory
they are consuming in the system.
Although those are memtable statistics, I am grouping them under the "cf_stats"
structure: being the column family a central piece of the puzzle, it is reasonable
to assume that a lot of metrics about it would be potentially welcome in the future.
Note that we don't want to reuse the "stats" structure in the column family: for once,
the fields not always map precisely (pending flushes, for instance, only tracks explicit
flushes), and also the stats structure is a lot more complex than we need.
Signed-off-by: Glauber Costa <glommer@scylladb.com>
* seastar 5c10d3e...20bf03b (5):
> do not re-throw exception to get to an exception pointer
> Adding timeout counter to the rpc
> configure.py: support for pkg-config before release 0.28
> future: don't forget to warn about ignored exception
> tutorial: continue network API section
Found by debug build
==10190==ERROR: AddressSanitizer: new-delete-type-mismatch on 0x602000084430 in thread T0:
object passed to delete has wrong type:
size of the allocated type: 16 bytes;
size of the deallocated type: 8 bytes.
#0 0x7fe244add512 in operator delete(void*, unsigned long) (/lib64/libasan.so.2+0x9a512)
#1 0x3c674fe in std::default_delete<dht::range_streamer::i_source_filter>::operator()(dht::range_streamer::i_source_filter*)
const /usr/include/c++/5.1.1/bits/unique_ptr.h:76
#2 0x3c60584 in std::unique_ptr<dht::range_streamer::i_source_filter, std::default_delete<dht::range_streamer::i_source_filter> >::~unique_ptr()
/usr/include/c++/5.1.1/bits/unique_ptr.h:236
#3 0x3c7ac22 in void __gnu_cxx::new_allocator<std::unique_ptr<dht::range_streamer::i_source_filter,
std::default_delete<dht::range_streamer::i_source_filter> > >::destroy<std::unique_ptr<dht::range_streamer::i_source_filter,
std::default_delete<dht::range_streamer::i_source_filter> > >(std::unique_ptr<dht::range_streamer::i_source_filter,
std::default_delete<dht::range_streamer::i_source_filter> >*) /usr/include/c++/5.1.1/ext/new_allocator.h:124
...
Fixes#549.
Being clinically absent-minded, aggregate query support (i.e. count(...))
was left out of the "paging" change set.
This adds repeated paged querying to do aggregate queries (similar to
origin). Uses "batched" paging.
Until the compaction manager api would be ready, its failing command
causes problem with nodetool related tests.
Ths patch stub the compaction manager logic so it will not fail.
It will be replaced by an actuall implementation when the equivelent
code in compaction will be ready.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
This patch adds a compaction info object and an API that returns it.
It will be mapped to the JMX getCompactions that returns a map.
The use of an object is more RESTFull and will be better documented in
the swagger definition file.
For compatibility reasons, compaction_strategy should accept both class
name strategy and the full class name that includes the package name.
In origin the result name depends on the configuration, we cannot mimic
that as we are using enum for the type.
So currently the return class name remains the class itself, we can
consider changing it in the future.
If the name is org.apache.cassandra.db.compaction.Name the it will be
compare as Name
The error message was modified to report the name it was given.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Fixes#545
"Slight file format change for commitlog segments, now incluing
a scylla "marker". Allows for fast-fail if trying to load an
Origin segment.
WARNING: This changes the file format, and there is no good way for me to
check if a CL is "old" scylla, or Origin (since "version" is the same). So
either "old" scylla files also fail, or we never fail (until later, and
worse). Thus, if upgrading from older to this patch ensure to
have cleaned out all commit logs first."
Fixes#355
"Implements query paging similar to origin. If driver sets a "page size" in
a query, and we cannot know that we will not exceed this limit in a single
query, the query is performed using a "pager" object, which, using modified
partition ranges and query limits, keeps track of returned rows to "page"
through the results.
Implementation structure sort of mimics the origin design, even though it
is maybe a little bit overkill for us (currently). On the other hand, it
does not really hurt.
This implementation is tested using the "paging_test" subset in dtest.
It passes all test except:
* test_paging_using_secondary_indexes
* test_paging_using_secondary_indexes_with_static_cols
* test_failure_threshold_deletions
The two first because we don't have secondary indexes yet, the latter
because the test depends on "tombstone_failure_threshold" in origin.
Potential todo: Currently the pager object does not shortcut result
building fully when page limit is exceeded. Could save a little work
here, but probably not very significant."
Allows us fail fast if someone tries to replay an Origin commit log.
WARNING: This changes the file format, and there is no good way for me to
check if a CL is "old" scylla, or Origin (since "version" is the same). So
either "old" scylla files also fail, or we never fail (until later, and
worse). Thus, if upgrading from older to this patch, likewise, ensure to
have cleaned out all commit logs first.
* Static query method to determine if paging might be required
(very conservative - almost all querys will be paged me thinks).
* Static factory method for pager
* Actual pager implementation
Pager object uses three variables to keep track of paging state:
1.) Last partition key - partition key of last partion processed
-> next partition to start process
2.) Last clustering key, i.e. row offset within last key partition,
i.e. how far we got last time
3.) Max remaining - max rows to process further, i.e. initial limit -
processed so far
Partition ranges are modified/removed so that we begin with "Last key",
if present. (Or end with, in the case of reversed processing)
A counting visitor then keeps count of rows to include in processing.
Basic interface for paging control objects.
We probably do not need virtual behaviour for paging, but on the other
hand it does not really cost much, and it keeps a nice symmetry with
origin.
Allows for having more than one clustering row range set, depending on
PK queried (although right now limited to one - which happens to be exactly
the number of mutiplexing paging needs... What a coincidence...)
Encapsulates the row_ranges member in a query function, and if needed holds
ranges outside the default one in an extra object.
Query result::builder::add_partition now fetches the correct row range for
the partition, and this is the range used in subsequent iteration.
Note: serial format blob is different compared to origin, due to scyllas
different internal architecture. I.e. we query actual rows.
But drivers etc ignore the content of the blob, it is opaque.
Currently, there are multiple places we can close a session, this makes
the close code path hard to follow. Remove the call to maybe_completed
in follower_start_sent to simplify closing a bit.
- stream_session::follower_start_sent -> maybe_completed()
- stream_session::receive_task_completed -> maybe_completed()
- stream_session::transfer_task_completed -> maybe_completed()
- on receive of the COMPLETE_MESSAGE -> complete()
nodetool decommission node 127.0.0.2, on node 127.0.0.1, I saw:
DEBUG [shard 0] gossip - failure_detector: Forcing conviction of 127.0.0.1
TRACE [shard 0] gossip - convict ep=127.0.0.1, phi=8, is_alive=1, is_dead_state=0
TRACE [shard 0] gossip - marking as down 127.0.0.1
INFO [shard 0] gossip - inet_address 127.0.0.1 is now DOWN
DEBUG [shard 0] storage_service - on_dead endpoint=127.0.0.1
This is wrong since the argument for send_gossip_shutdown should be the
node being shutdown instead of the live node.
Since the introduction of sets::element_discarder sets::discarder is
always given a set, never a single value.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Currently sets::discarder is used by both set difference and removal of
a single element operations. To distinguish between them the discarder
checks whether the provided value is a set or something else, this won't
work however if a set of frozen sets is created.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Error handling in column_family::try_flush_memtable_to_sstable() is
misplaced. It happens after update_cache(), so writing sstable may
have succeeded, but moving memtable into the cache may have failed.
update_cache() destroys memtable even if it fails, but error handler
is not aware of it (it does not even distinguish whether error happened
during sstable creation or moving into cache) and when it tells caller
to retry it retries with already destroyed memtable. Fix it by ignoring
moving to cache errors.
This reverts commit fff37d15cd.
Says Tomek (and the comment in the code):
"update_cache() must be called before unlinking the memtable because cache + memtable at any time is supposed to be authoritative source of data for contained partitions. If there is a cache hit in cache, sstables won't be checked. If we unlink the memtable before cache is updated, it's possible that a query will miss data which was in that unlinked memtable, if it hits in the cache (with an old value)."
Error handling in column_family::try_flush_memtable_to_sstable() is
misplaced. It happens after update_cache(), so writing sstable may
have succeeded, but moving memtable into the cache may have failed.
update_cache() destroys memtable even if it fails, but error handler
is not aware of it (it does not even distinguish whether error happened
during sstable creation or moving into cache) and when it tells caller
to retry it retries with already destroyed memtable. Fix it by ignoring
moving to cache errors.
nodetool decommission hangs forever due to a recursive lock.
decommission()
with api lock
shutdown_client_servers()
with api lock
stop_rpc_server()
with api lock
stop_native_transport()
Fix it by calling helpers for stop_rpc_server and stop_native_transport
without the lock.
std::set_difference requires the container to be sorted which is not
true here, use remove_if.
Do not use assert, use throw instead so that we can recover from this
error.
Currently error code is attached to a future returned by when_all() which
is never is exceptional one, but it may hold exceptional future as a
first element. Move error handling close to where error it tries to
catch is generated instead.
Let's move the code that prints that a compaction succeeded only
after the code that catches exception on either read or write
fibers. Let's also get rid of done and use repeat instead in
the read fiber.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
If write timeout and last acknowledgement needed for CL happen simultaneously
_ready can be sent to be exceptional by the timeout handler, but since
removal of the response handler happens in continuation it may be
reordered with last ack processing and there _ready will be set again
which will trigger assert. Fix it by removing the handler immediately,
no need to wait for continuation. It makes code simpler too.
The get_cm_stats gets a pointer to a field in the stats object. It
should capture it by value or segmentation falut may occure when the
caller gets out of scope.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Currently, we don't let the user know even what is the filename that failed.
That information should be included in the message.
Signed-off-by: Glauber Costa <glommer@scylladb.com>
This assert (in write fiber) would fail if read fiber failed
because the variable done will not be set to true.
The use of assert is very bad, because it prevents scylla
from proceeding, which is possible.
To solve it, let's trigger an exception if done is not true.
We do have code that will wait for both read and write fibers,
and catch exceptions, if any.
Closes#523.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Since 4641dfff24, query_state keeps a
copy of client_state, not a reference. Therefore _cl is no longer
updated by queries using _qp. Fix by using the client_state from _qp.
Fixes#525.
All responses sent from the server have protocol version set to
connection::_version which is set to the version used by the client
in its first message. However, if the protocol version used by the
client is unuspported or invalide the server should use the latest
version it recognizes.
This solves problem with version negotiation with Java driver. The
driver first sends a request in the latest version it recognizes, if
that fails it retries with the version that server has used in the error
message. If that fails as well it gives up. However, since Scylla always
responds with the same version that the client has used the negotiation
always fails if the client supports more protocol version than the
server.
Refs #317.
Signed-off-by: Paweł Dziepak <pdziepak@scylladb.com>
Get initial tokens specified by the initial_token in scylla.conf.
E.g.,
--initial-token "-1112521204969569328,1117992399013959838"
--initial-token "1117992399013959838"
It can be multiple tokens split by comma.
"This series adds the missing functionality that the nodetool describering would work.
It import the missing functionality from origin.
After this patch the API:
GET /storage_service/describe_ring/{keyspace}
will be available"
This patch chanages the API to support describe ring instead of describe
ring jmx that will be implemented in the jmx server.
The API will return a list of objects instead of string.
An additional api was added as the equivelent to the jmx call with an
empty param.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
This patch adds the following methods implementation:
getRpcaddress
getRangeToAddressMap
getRangeToAddressMapInLocalDC
describeRing
getAllRanges
Those methods are used as part of the describe_ring method
implementation.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
The storage server uses the token_range in origin to return inforamtion
about the ring.
This import the structures. The functionality in origin is redundant in
this case and was not imported.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Use all the disks except the one for rootfs for RAID0 which stores
scylla data. If only one disk is available warn the user since currently
our AMI's rootfs is not XFS.
[fedora@ip-172-31-39-189 ~]$ cat WARN.TXT
WARN: Scylla is not using XFS to store data. Performance will suffer.
Tested on AWS with 1 disk, 2 disks, 7 disk case.
(cherry picked from commit 49d6cba471)
Mistakenly didn't included on yum repository for AMI patchset, but it's needed
Signed-off-by: Takuya ASADA <syuu@cloudius-systems.com>
(cherry picked from commit 8587c4d6b3)
The nodetool cleanup command is used in many of the tests, because the
API call is not implemented it causes the tests to fail.
This is a workaround until the cleanup will be implemented, the method
return successfuly.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
Normally an API call that is not implemented should fail, there are
cases that as a workaround an API call is stub, in those cases a warning
is added to indicate that the API is not implemented.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
This patch do the following:
It adds a getter for the completed respond messages (i.e. the total
messages that were sent by the server)
It replaces the return mapping for the statistics to use the key, value
notation that is used in the jmx side.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
This adds the read repair statistics to he storage_proxy stats and adds
to its implementation incrementing the counters value.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
The API needs to get the stats from the rpc server, that is hidden from the
messaging service API.
This patch adds a foreach function that goes over all the server stats
without exposing the server implementation.
Signed-off-by: Amnon Heiman <amnon@scylladb.com>
"The main objective of the series is to introduce statistics about ongoing
read/writes and especially those that are done in the background (acknowledged,
but uncompleted), but it contains some cleanups as well."
Add statistics for ongoing reads and ongoing background reads. Read is
a background one if it was acknowledged, but there still work to do to
complete it.
"Commit 4cd9c4c0c5441cf55e280c6f2f2e5529426b9c98 introduced a minor
issue: a wrong snitch instance may be used when updating a Gossiper state
(if I/O CPU is different from CPU0).
In order to fix this issue a local snitch instance on CPU0 should be used,
just like a Gossiper local instance.
We have to move some interfaces to i_endpoint_snitch
from being private in a gossiping_property_file_snitch in order to be
able to access it using snitch_ptr handle."
Don't ignore yet another returned future in reload_configuration().
Since commit 5e8037b50a
storage_service::gossip_snitch_info() returns a future.
This patch takes this into an account.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
When we access a gossiper instance we use a _gossip_started
state of a snitch, which is set in a gossiper_starting() method.
gossiper_starting() method however is invoked by a gossiper on CPU0
only therefore the _gossip_started snitch state will be set for an
instance on CPU0 only.
Therefore instead of synchronizing the _gossip_started state between
all shards we just have to make sure we check it on the right CPU,
which is CPU0.
This patch fixes this issue.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Adjust the interface and distribution of prefer_local parameter read
from a snitch property file with the rest of similar parameters (e.g. dc and rack):
they are read and their values are distributed (copied) across all shards'
instances.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Make reload_gossiper_state() be a virtual method
of a base class in order to allow calling it using a snitch_ptr
handle.
A base class already has a ton of virtual methods so no harm is
done performance-wise. Using virtual methods instead of doing
dynamic_cast results in a much cleaner code however.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Move the member and add an access method.
This is needed in order to be able to access this state using
snitch_ptr handle.
This also allows to get rid of ec2_multi_region_snitch::_helper_added
member since it duplicates _gossip_started semantics.
Signed-off-by: Vlad Zolotarov <vladz@cloudius-systems.com>
* seastar 9ae6407...258daf9 (6):
> rpc server: Add pending and sent messages to server
> scripts: posix_net_conf.sh: Use a generic logic for RPS configuring
> scripts: posix_net_conf.sh: allow passing a NIC name as a parameter
> doc: link to the tutorial
> tutorial: begin documenting the network API
> slab: remove bogus uintptr_t definition
"In 5e8037b50a (gossip: Futurize
add_local_application_state()) , we futurized add_local_application_state.
However, not all of the callers are futurized. Fix it up."
"- Fix snitch names from EC2XXX to Ec2XXX to align with configuration.
- Copy cassandra-rackdc.properties file to /var/lib/scylla/conf
- Set SCYLLA_HOME before booting process"
During testing build, the debugging statement at the end
of the function body (after return statements) causes compilation to
fail due to the flag -Werror=return-type:
service/storage_service.cc: In member function ‘future<> service::storage_service::clear_snapshot(sstring, std::vector<basic_sstring<char, unsigned int, 15u> >)’:
service/storage_service.cc:1358:1: error: control reaches end of non-void function [-Werror=return-type]
Which traces back to 21f84d77. Let's attach a then_wrapped()
clause to parallel_for_each() adding the debug message as
suggested by Avi.
CC: Glauber Costa <glommer@scylladb.com>
Signed-off-by: Lucas Meneghel Rodrigues <lmr@scylladb.com>
We are ignoring the future returned by seastar::async. Futurize it so
caller can wait for the application state to be actually applied.
In addition, dropping the unused add_local_application_states function.
We use boost::any to convert to and from database values (stored in
serlialized form) and native C++ values. boost::any captures information
about the data type (how to copy/move/delete etc.) and stores it inside
the boost::any instance. We later retrieve the real value using
boost::any_cast.
However, data_value (which has a boost::any member) already has type
information as a data_type instance. By teaching data_type intances about
the corresponding native type, we can elimiante the use of boost::any.
While boost::any is evil and eliminating it improves efficiency somewhat,
the real goal is growing native type support in data_type. We will use that
later to store native types in the cache, enabling O(log n) access to
collections, O(1) access to tuples, and more efficient large blob support.
"gossiping_property_file_snitch checks its property
file (cassandra-rackdc.properties) for changes every minute and
if there were changes it re-registers the helper and initiates
re-read of the new DC and Rack values in the corresponding places.
Therefore we need the ability to unregister/register the corresponding subscriber
at the same time when a subscriber list is possibly iterated by
some other asynchronous context on the current CPU.
The current gossiper implementation assumes that subscribers list may not be
changed from the context different from the one that iterates on their list.
So, this had to be fixed.
There was also missing an update_endpoint(ep) interface in the locator::topology
class and the corresponding token_metadata::update_topology(ep) wrapper.
Also there were some bugs in the gossiping_property_file::reload_configuration()
method."
On hindsight, it doesn't make much sense to print an
empty string, so let's only print stdout if it's non
None, non empty.
Signed-off-by: Lucas Meneghel Rodrigues <lmr@scylladb.com>
Do not hold the api lock while streaming the data since it might take a
long time, so we need to reconcile other operations while we are in the
middle of rebuild.
Use class or struct similar to the object you need the serializer for.
Use namespace when applicable.
##keywords
* class/struct - a class or a struct like C++
class/struct can have final or stub marker
* namespace - has the same C++ meaning
* enum class - has the same C++ meaning
* final modifier for class - when a class mark as final it will not contain a size parameter. Note that final class cannot be extended by future version, so use with care
* stub class - when a class is mark as stub, it means that no code will be generated for this class and it is only there as a documentation.
* version attributes - mark with [[version id ]] mark that a field is available from a specific version
* template - A template class definition like C++
##Syntax
###Namespace
```
namespace ns_name { namespace-body }
```
* ns_name: either a previously unused identifier, in which case this is original-namespace-definition or the name of a namespace, in which case this is extension-namespace-definition
* namespace-body: possibly empty sequence of declarations of any kind (including class and struct definitions as well as nested namespaces)
* identifier: the name of the enumeration that's being declared.
* enum-base: colon (:), followed by a type-specifier-seq that names an integral type (see the C++ standard for the full list of all possible integral types).
* enumerator-list: comma-separated list of enumerator definitions, each of which is either simply an identifier, which becomes the name of the enumerator, or an identifier with an initializer: identifier = integral value.
Note that though C++ allows constexpr as an initialize value, it makes the documentation less readable, hence is not permitted.
* type: Any valid C++ type, following the C++ notation. note that there should be a serializer for the type, but deceleration order is not mandatory
* member-access: is the way the member can be access. If the member is public it can be the name itself. if not it could be a getter function that should be followed by braces. Note that getter can (and probably should) be const methods.
* attributes: Attributes define by square brackets. Currently are use to mark a version in which a specific member was added [ [ version version-number] ] would mark that the specific member was added in the given version number.
###template
`template < parameter-list > class-declaration`
* parameter-list - a non-empty comma-separated list of the template parameters.
* class-decleration - (See class section) The class name declared become a template name.
##IDL example
Forward slashes comments are ignored until the end of the line.
"description":"If the value is the string 'true' with any capitalization, repair only the first range returned by the partitioner.",
"required":false,
"allowMultiple":false,
"type":"string",
"paramType":"query"
},
{
"name":"parallelism",
"description":"Repair parallelism, can be 0 (sequential), 1 (parallel) or 2 (datacenter-aware).",
"required":false,
"allowMultiple":false,
"type":"string",
"paramType":"query"
},
{
"name":"incremental",
"description":"If the value is the string 'true' with any capitalization, perform incremental repair.",
"required":false,
"allowMultiple":false,
"type":"string",
"paramType":"query"
},
{
"name":"jobThreads",
"description":"An integer specifying the parallelism on each node.",
"required":false,
"allowMultiple":false,
"type":"string",
"paramType":"query"
},
{
"name":"ranges",
"description":"An explicit list of ranges to repair, overriding the default choice. Each range is expressed as token1:token2, and multiple ranges can be given as a comma separated list.",
"required":false,
"allowMultiple":false,
"type":"string",
"paramType":"query"
},
{
"name":"columnFamilies",
"description":"Which column families to repair in the given keyspace. Multiple columns families can be named separated by commas. If this option is missing, all column families in the keyspace are repaired.",
"required":false,
"allowMultiple":false,
"type":"string",
"paramType":"query"
},
{
"name":"dataCenters",
"description":"Which data centers are to participate in this repair. Multiple data centers can be listed separated by commas.",
"required":false,
"allowMultiple":false,
"type":"string",
"paramType":"query"
},
{
"name":"hosts",
"description":"Which hosts are to participate in this repair. Multiple hosts can be listed separated by commas.",
"required":false,
"allowMultiple":false,
"type":"string",
"paramType":"query"
},
{
"name":"trace",
"description":"If the value is the string 'true' with any capitalization, enable tracing of the repair.",
"required":false,
"allowMultiple":false,
"type":"string",
@@ -1945,6 +2028,20 @@
}
}
},
"map_string_double":{
"id":"map_string_double",
"description":"A key value mapping between a string and a double",
"properties":{
"key":{
"type":"string",
"description":"The key"
},
"value":{
"type":"double",
"description":"The value"
}
}
},
"maplist_mapper":{
"id":"maplist_mapper",
"description":"A key value mapping, where key and value are list",
throwexceptions::invalid_request_exception(sprint("Cannot rename unknown column %s in table %s",from,column_family()));
}
if(schema->get_column_definition(to->name())){
throwexceptions::invalid_request_exception(sprint("Cannot rename column %s to %s in table %s; another column of that name already exist",from,to,column_family()));
}
if(def->is_part_of_cell_name()){
throwexceptions::invalid_request_exception(sprint("Cannot rename non PRIMARY KEY part %s",from));
}
if(def->is_indexed()){
throwexceptions::invalid_request_exception(sprint("Cannot rename column %s because it is secondary indexed",from));
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
