MariaDB 10.6.18 stuck in restart loop
In the past month, we have observed two instances where our Maria DB instance crashed, and then entered an endless restart loop.
In both cases the issue seemed related to Maria DB's "doublewrite buffer".
MariaDB never recovers from the crash and restarts again 3-4 minutes later.
The only solution was to restore a previous snapshot from before the database entered the corrupt state.
We suspect that upgrading to at least 10.11.44 will mitigate this, as we know that there were significant changes to the doublewrite buffer.
1. Can anyone confirm whether this upgrade will make a difference? 2. Does MariaDB allow for a "safe mode" of some kind that would let the instance start without attempting to recover from a previous crash?
Note that this database is hosted within AWS.
Thank you for any suggestions.
Answer Answered by Ross Metcalf in this comment.
To give an update to others in the future who might face a similar issue, and aren't MariaDB experts -
In the end, this issue had nothing to do with the "doublewrite buffer". Or, signs pointing to the doublewrite buffer were misleading.
The real issue was related to a "spatial index" which had become corrupt.
Dropping and re-creating the index fixes the issue.
