scandium has issues. It's triggered twice an hour by cocoplum, and quite often that trigger hangs for >10 minutes. This seriously gets in the way of fixing bug 36535; I think we could safely move to 30-minute publisher cycles now if not for this problem. IMO we should trigger it in the background so that this isn't quite such a problem.
<cjwatson> lamont: the trigger of scandium from cocoplum seems sporadically very slow at the moment; at 08:47 it took 16 minutes, and at 13:04 it took 11 minutes. Can you look and see if anything's up?
<elmo> cjwatson: this is the ssh packets being lost thing
<elmo> cjwatson: I'm hoping/praying it's due to scandium being hardy
<cjwatson> I'm wondering if it would make sense to background that trigger from LP
<elmo> cjwatson: scandium is due to be moved DCs early next year and it's replacement won't/shouldn't have this problem
<elmo> cjwatson: I think that's a reasonable workaround for now, sure
<cjwatson> because, if not for this, I'm pretty sure that we could switch to 30-minute publisher cycles now
<elmo> at least for anything that triggers regularly
<cjwatson> cocoplum triggers scandium twice an hour and blocks on it
<elmo> cjwatson: yeah, sure, background away
<cjwatson> elmo: OK, I'll see if I can get a branch landed for that, thanks
Fixed in stable r14538 <http://