Package: mariadb-server Version: 1:10.3.27-0+deb10u1
Mariadb aborts an long semaphore list relating to analyze table commands for large databases. After I ran analyze on a big table on a server with slow storage, I aborted the command and restarted the server. Since then it seems that mariadb crashes every day at some time with a long semaphore list. The last crash was monitored, and I stopped the workload on the db. The only queries still running where 12 insert and updates that just did never commit. These queries are usual application queries that hit the db hundred times per minute. After searching it seems that this is fixed upstream with https://jira.mariadb.org/browse/MDEV-24275 and has to do with "InnoDB persistent stats analyze forces full scan forcing lock crash". We are running Debian buster 10.8 and I think this should be fixed as this is the only mariadb version available for stable. Mariadb logs [ERROR] [FATAL] InnoDB: Semaphore wait has lasted > 600 seconds. We intentionally crash the server because it appears to be hung. Kind regards
