Springle has submitted this change and it was merged. Change subject: Reduce MariaDB thread_pool_stall_limit to 100ms ......................................................................
Reduce MariaDB thread_pool_stall_limit to 100ms Some hard-to-pin-down combination of the Mediawiki load balancing code, our low connect timeout, the tcp back_log, and the thread pool causes bursts of EINTR connection failures in dberror.log. One early symptom is many connections with 'unauthenticated user' sitting in State=login. A later one is connections stampeding between slaves in a kind of feedback loop. Reducing the stall limit seems to help, but this needs further attention. Change-Id: Ic464000c6aebfc32c2eb994b76c6d2f903b28a02 --- M templates/mariadb/production.my.cnf.erb 1 file changed, 1 insertion(+), 0 deletions(-) Approvals: Springle: Looks good to me, approved jenkins-bot: Verified diff --git a/templates/mariadb/production.my.cnf.erb b/templates/mariadb/production.my.cnf.erb index 46f637f..f3767ba 100644 --- a/templates/mariadb/production.my.cnf.erb +++ b/templates/mariadb/production.my.cnf.erb @@ -25,6 +25,7 @@ temp-pool thread_handling = pool-of-threads +thread_pool_stall_limit = 100 max_connections = 5000 back_log = 500 extra_max_connections = 10 -- To view, visit https://gerrit.wikimedia.org/r/192254 To unsubscribe, visit https://gerrit.wikimedia.org/r/settings Gerrit-MessageType: merged Gerrit-Change-Id: Ic464000c6aebfc32c2eb994b76c6d2f903b28a02 Gerrit-PatchSet: 1 Gerrit-Project: operations/puppet Gerrit-Branch: production Gerrit-Owner: Springle <[email protected]> Gerrit-Reviewer: Springle <[email protected]> Gerrit-Reviewer: jenkins-bot <> _______________________________________________ MediaWiki-commits mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/mediawiki-commits
