jcrespo added a comment. Actually, now that I see, both could be responsible, as both are reported around 15:30 UTC.
> Given we have statement based replication Not 100% truth everywhere anymore. See our puppet config. > these should show up for the slaves as well, in case they run for longer > periods of time, shouldn't they? No, I do not think our monitoring system would catch system user queries on slaves even if we used exclusively STATEMENT replication, and even if it would, that would not discard having problems with them because it takes them only being <10 seconds to not be noticed but create pileups due to single-thread replication. The fact that this created lag on all 5 of our slaves (usually) means that the problem is in the centralized distribution point of transactions (master)- unless it is a progressive failover where there are depooled one by one. TASK DETAIL https://phabricator.wikimedia.org/T123867 EMAIL PREFERENCES https://phabricator.wikimedia.org/settings/panel/emailpreferences/ To: jcrespo Cc: Aklapper, StudiesWorld, aaron, daniel, aude, Lydia_Pintscher, Multichill, jcrespo, hoo, Wikidata-bugs, Mbch331, Krenair _______________________________________________ Wikidata-bugs mailing list Wikidata-bugs@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikidata-bugs