[04:20] <tsimonq2> When does publisher maintenance typically start? I've heard 6 AM UK time on Sundays but nothing definitive.
[09:20] <cjwatson> tsimonq2: 59 05  * * 0 /srv/launchpad.net/production/launchpad/cronscripts/publishing/cron.daily-ppa >> /srv/launchpad.net/production-logs/lp_publish/cron.ppa.log 2>&1
[09:21] <cjwatson> One of our more impressively misnamed scripts
[09:35] <tsimonq2> cjwatson: Aha, thanks.
[09:37] <tsimonq2> cjwatson: Wait, so despite the comment in the script and the script naming, this thing is done weekly?
[09:37] <tsimonq2> Hm, looks like it.
[09:37] <tsimonq2> I'd be curious to hear the rationale for that.
[09:47] <cjwatson> tsimonq2: Presumably because running a thing daily that takes several hours is a bit rude.
[09:47] <cjwatson> And nobody ever got round to renaming the thing.
[09:47] <cjwatson> It was probably initially daily.
[09:48] <cjwatson> (And also it should eventually be removed once our publishing machinery is better - it's a workaround.)
[09:54] <tsimonq2> Fair.
[10:12] <tsimonq2> I wonder if that could be sped up if at all possible... I'd imagine the bottleneck is find so the output could be piped to a temp file and then rmdir could run at once?
[10:12] <tsimonq2> (So, get the lock after find runs.)
[10:23] <cjwatson> I doubt that would help measurably.
[10:23] <cjwatson> And the reason this blocks publishing is that it would otherwise be racy.
[10:24] <cjwatson> The correct fix is to chase down all the places where the publisher doesn't do this sort of empty-directory cleanup for itself, I think.  (Or it would become irrelevant if we ever do diskless PPAs.)