0

Our Mariadb database (1To) encounter slow down during some minutes but we don't find the root cause : there no CPU or IO overhead but locks OS dans lot of process taking time on sending data and statistics.

This happen one time day not at same hour.

enter image description here

Update 1 : There are lot a slow query during the slow down not easy to find the one that cause that. Slow query graph :

enter image description here

Update 2 : When we SHOW PROCESS list we see lot of queries in "sending data", all query are different, it seems to be normal query that are slow for some minutes.

Update 3 : I found in InnoDB Status a long transaction running, I believe that can be the root cause :
enter image description here

Improve this question
9
  • 2
    I'd enable the slow query log. It looks like a odd query doing a large amount of data fetching and should show up pretty clearly when it does occur.
    – danblack
    Feb 2 at 9:24
  • Is that on a Cloud service? When are backups performed? "1To" is a lot of "Sending data". Next time it happens, connect as root and capture SHOW PROCESSLIST;.
    – Rick James
    Feb 2 at 19:21
  • If you don't see much in the slowlog, lower long_query_time from its nearly-useless 10 second default.
    – Rick James
    Feb 2 at 19:22
  • Yes long_quert_time is setup to 10 sec. Because we thought that there ware a bottleneck in slowquerylog file.
    – Mr_Thorynque
    Feb 3 at 9:19
  • No backup was performed, there is a replicat for that.
    – Mr_Thorynque
    Feb 3 at 13:51

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.