Details
-
Bug
-
Status: Confirmed (View Workflow)
-
Critical
-
Resolution: Unresolved
-
10.5.21, 10.5.24, 10.6, 10.11, 11.4
-
Debian 11
Description
We have a system with fairly hight amount of deadlocking sessions. At some point (with no apparent precursor like application change or increase of load) it started reporting broken crash recovery (as if a redo log was overflown / overwritten), which results in an unusable database state.
2024-03-13 8:33:35 79392757 [ERROR] InnoDB: Crash recovery is broken due to insufficient innodb_log_file_size; last checkpoint LSN=13565873821532, current LSN=13581335735451. Shutdown is in progress.
|
This happened on 2 occasions, first with version 10.5.21 and later on 10.5.24 too. Synced replica after failover succumbed to the same issue, not immediately, but within hour(s) in both cases as well.
In the second case, redo has been increased to 16G with a wishful thinking it might prevent further issues (even though redo should be flushed to disk with a priority at certain thresholds 75/~90? if I am not wrong).
Error log is modified (removed sensitive information, user disconnect and semaphore wait flood).
Attachments
Issue Links
- relates to
-
MDEV-14462 Confusing error message: ib_logfiles are too small for innodb_thread_concurrency=0
- Closed
-
MDEV-21010 Mariadb hangs (during a backtrace), stops responding to new connections
- Open
-
MDEV-27774 Reduce scalability bottlenecks in mtr_t::commit()
- Closed
-
MDEV-29982 Improve the InnoDB log overwrite error message
- Closed
-
MDEV-34830 LSN in the future is not being treated as serious corruption
- Closed