doc: add troubleshooting for failed schema sync

Fixes https://github.com/scylladb/scylladb/issues/12133

This commit adds a Troubleshooting article to support
users when schema sync failed on their cluster.

Closes #13709
This commit is contained in:
Anna Stuchlik
2023-04-28 12:12:41 +02:00
committed by Botond Dénes
parent 1a3545b13d
commit 4898a20ae9
2 changed files with 18 additions and 0 deletions

View File

@@ -10,6 +10,7 @@ Cluster and Node
Node Joined With No Data </troubleshooting/node-joined-without-any-data>
SocketTimeoutException </troubleshooting/nodetool-memory-read-timeout/>
NullPointerException </troubleshooting/nodetool-nullpointerexception/>
Failed Schema Sync </troubleshooting/failed-schema-sync/>
.. raw:: html
@@ -25,6 +26,7 @@ Cluster and Node
* :doc:`Node Joined With No Data </troubleshooting/node-joined-without-any-data>`
* :doc:`Nodetool fails with SocketTimeoutException 'Read timed out' </troubleshooting/nodetool-memory-read-timeout>`
* :doc:`Nodetool Throws NullPointerException </troubleshooting/nodetool-nullpointerexception>`
* :doc:`Failed Schema Sync </troubleshooting/failed-schema-sync>`
.. raw:: html

View File

@@ -0,0 +1,16 @@
Failed Schema Sync
===================
The total binary footprint of all ScyllaDB columns, tables, etc., must fit a single commit log segment size
divided by two. If this requirement is not met, schema sync may fail. The error message may resemble the following:
.. code:: console
Oct 06 22:12:47 ip-172-21-1-122 scylla[1213356]: [shard 0] storage_service - Fail to pull schema from 172.21.2.210: std::invalid_argument (Mutation of 27837439 bytes is too large for the maximum size of 16777216)
Oct 06 22:12:47 ip-172-21-1-122 scylla[1213356]: [shard 0] migration_manager - Pulling schema from 172.21.2.210:0
In such a case, you may need to increase the commitlog segment size limit. To do this,
go to ``/etc/scylla/scylla.yaml`` and set the ``commitlog_segment_size_in_mb`` parameter to a higher value.
.. note::
The ``commitlog_segment_size_in_mb`` parameter must be set to the same value on **all nodes** in a cluster.