Asher Feldman wrote:
Just a heads up that the last of the 1.19 migrations, to add the sha1 column to enwiki.revision is going to be running throughout this week. Don't be alarmed by replication lag messages for s1 dbs in irc. I'm going to juggle which db watchlist queries go to during the migration, so nothing should be noticeable on the site.
Hi.
This is causing significant problems for the scripts that I run that rely on the MediaWiki API. The scripts were designed to be respectful of database lag, but with db36 reporting 20879 seconds of lag (and rising), the scripts are all getting hung as they patiently wait for lag to go back down.
I'm not sure of the exact configuration, but it seems like nearly every API request is being handled by the lagged server (db36)? Or perhaps my scripts just have terrible luck.
In any case, is it possible to depool the server or do something to mitigate this? I don't think we want to encourage script writers to write scripts that simply ignore the lag, but there aren't many other options when it's outrageously high. Alternately, could the other servers be assigned to API duty?
MZMcBride