we had a tree closure lasting over a few hours due to aws pending jobs (compile/build jobs) not finding a valid spot. Theories included spot bidding prices being too high, not enough ip's in the AZ subnet group, and not enough slave names for non jacuzzi builders. Patches were put in place for the first two theories but it seems like only upping the ondemand limit actually solved the issue. This is of course is not a solution so we should look at adding more slave names to our non-jacuzzi builders. Apparently we added locked a lot of slaves to our jacuzzis in the the last week and this closure may be the result of that bug. see https://bugzilla.mozilla.org/show_bug.cgi?id=1027308#c16 for more details
also rail, we added a subnet that appeared missing in our config for east-1c: https://hg.mozilla.org/build/cloud-tools/rev/ec66195a91fd and, we should probably revert these two band aid fixes once we fix this bug: https://hg.mozilla.org/build/cloud-tools/rev/eae3f6598284 https://hg.mozilla.org/build/cloud-tools/rev/64155873112f
Created attachment 8442882 [details] [diff] [review] spots.diff Once this is in production I'll need to add the slave to slavealloc, that's it.
Comment on attachment 8442882 [details] [diff] [review] spots.diff https://hg.mozilla.org/build/buildbot-configs/rev/a51f0c4920c4
(In reply to Rail Aliiev [:rail] from comment #5) > Created attachment 8442900 [details] > spot-bld.csv > > To be added to slavealloc Done
buildbot-config patch live in production :)