[15:25:31] gwicke: So, regarding references [15:25:34] https://phabricator.wikimedia.org/T130663 [16:39:58] legoktm: Do you know of any bugs left that block converting CentralAuth to extension registration? (cc tgr) [16:51:25] Krinkle: the event is coming up, so lets see if they come back with ambitious plans? [17:30:57] anomie: the sessionmanager conditionals can't be converted [17:32:04] legoktm: We couldn't use a callback for that? Although if we don't care about non-master BC we can make those unconditional now. [17:34:08] legoktm: The current issue is that we want to use the callback to do an AuthManager conditional ;) [17:36:01] callback would work [17:36:15] just a bit ugly [17:56:16] getting a bunch of "could not allocate 218103807 bytes for translation cache" errors when running vagrant git-update today after a week [17:56:45] https://gist.github.com/Krinkle/e835a3e43c0238502febb66af872ff41 [17:57:21] hmmm... hhvm unhappy? [17:58:36] https://github.com/facebook/hhvm/blob/50e7cddb4de7a61d1090a8529b22aae31e712aa5/hphp/runtime/vm/jit/code-cache.cpp#L133-L137 [18:03:49] Filed https://phabricator.wikimedia.org/T134912 [18:03:58] it pauses for several minutes before outputting that [18:04:14] Each time [18:04:45] well the pause is because the gti fetches run in parallel and only output at the end of each batch I think [18:04:49] *git [18:05:27] the weird thing about TC cache exhaustion is that it is generally seen on long running HHVM processes, not scripts [18:08:54] I wonder if we are trying to run too many things in parallel for the ram allocated to the VM? [18:09:18] We spawn one proc per cpu core in that parallel fetch I think [18:09:32] Krinkle: how much ram do you have allocated to your VM? [18:20:37] bd808: Whatever the default is? [18:21:10] bd808: 40GB storage, 1538 MB RAM [18:21:15] According to VirtualBox [18:23:05] bumped to 2048 [18:30:32] looks like 1538 is the default [18:30:42] hardcoded in mediawiki-vagrant definitions [18:48:59] Krinkle: max(1536, system ram / 4) is the default. for me that ends up with 2G to the vm [18:49:15] bd808: We need to get you some more RAM then. :-) [18:49:51] James_F: heh. my personal laptops both have 16G [18:50:17] the WMF was cheap 3 years ago and wouldn't spring for a ram upgrade for my work laptop [18:50:27] bd808: 32+ or go home. ;-) [18:55:05] I'd rather have more vms in Labs honestly ;) [19:04:23] bd808: I have 16GB ram [19:04:32] bd808: only cpu defaults to :auto [19:04:39] bd808: vagrant_ram default is not :auto [19:04:59] hmmm you are correct [19:05:13] maybe the max() thing is supposed to apply, but it isn't [19:06:03] it would apply but only after you set default to auto with `vagrant config vagrant_ram auto` [19:06:44] I'm vaguely remembering discussing this a couple years ago [19:07:23] * Krinkle is fine with 2GB and doesn't want to allocate 4GB to MW-V [19:07:34] Ah, vagrant config is nice [19:07:38] * Krinkle edits it by hand [19:07:49] This also explains why editing it in VirtualBox didn't work [19:07:52] vagrant changes it back [19:07:58] yeah [23:44:44] dapatrick: what does SDLC stand for? (https://www.mediawiki.org/wiki/Wikimedia_Security_Team/SDLC) [23:45:22] legoktm: software development life cycle. [23:45:33] I'll update the page to note that. [23:45:51] ah, thanks :) [23:46:16] No prob.