Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Won't Fix
-
5.5.50-galera
-
None
-
3 node cluster with Debian 7, mariadb 5.5.50 and galera 25.3.16
Description
After a while the following things happen:
1) mysql -e"show status;"
Hangs forever and never returns
2) Using the mysql cmd line client to run any sql query that writes to a DB either:
Hangs forever
or
Returns after 50 seconds with a lock timeout
3) Running sudo service mysql status
Shows a HUGE number of threads (over 20,000 for instance) and the number keeps rising
4) The general query log still shows all the queries being logged, including the "show status" and sql writes that hang forever/timeout after 50 secs.
5) The only way to "fix" it is to kill -9 mysqld and mysqld_safe and restart mysql.
This happens around once a day on these nodes.
I appreciate this is probably not an easy thing to fix based on this info, so if you can provide me with a bunch of commands to run to aide in debugging the next time it happens then I can add those commands to the nodes and when they detect mysql is in this state they can get the info you require.