[16:06:26] ^d: I want to backport multiple changes to WikimediaMessages and ContactPage to wmf4. Is just updating both extensions to HEAD in the release branch a bad idea? [[How to deploy code]] seems to think so but ... [16:06:45] <^d> I have no qualms with doing that [16:06:55] <^d> (others might?) I didn't write how to deploy code [16:07:09] cool. I'll see if I'm smart enough to make the backports [16:07:25] * bd808 fumbles with submodules [16:17:13] ^d: double checking here: the instructions on [[How to deploy code]] say to `git merge --ff-only` but that will not work because the release branch has an extra commit that changed .gitreview. I think the right thing to do is merge without ff-only and update the generated merge commit to explain why I'm pulling in +8 upstream commits. [16:26:40] <^d> bd808: --ff-only is pretty much wrong in the context of deployments :) [16:26:56] yeah. I thought so [16:27:41] ok one last question. When I have it all coommitted locally do I just push to origin wmf/1.26wmf4 to update gerrit? [16:48:49] ^d: `git push origin wmf/1.26wmf4` to update gerrit when I have all the right things in my local repo? [16:55:13] <^d> bd808: I think? [16:55:32] ^d: :) K. I'll give it a shot [16:55:44] I just didn't want to do the completely wrong thing [16:58:41] ^d: grumble. I don't have enough rights to push. "To push into this reference you need 'Push' rights." [16:58:58] I guess I can just cherry-pick all the damn patches [17:57:59] bd808: you can grant yourself the right temporarily [17:58:05] (or i can, if you lack the requisite permissions) [17:58:22] ori: I just did the checrry-picks [17:58:27] It's all good now [17:58:28] * ori nods [20:09:34] here's a good newbie question that folks can flame... [20:11:16] OMG WHAT A STUPID QUESTION LOL [20:11:21] sorry, go ahead. [20:11:23] (related to wmf deployment, not core) [20:11:25] hypothetically (!) if someone were going to write a service that would extract data from phabricator and do "cool stuff" with it, and this service were going to run on wikimedia servers, would any of these potential implementation languages either be (a) greeted with enthusiasm, or (b) chased away with pitchforks? Java, Ruby, Lua [20:12:22] any of those three would be fine, you are most likely not going to get pushback on language choice as long as it's principled [20:14:06] ori: cool, thanks. I think "expedient" would be more a factor than "principled" in this (hypothetical) case [20:17:09] meeple27: is this going to run in prod? [20:17:17] or on labs? [20:55:35] yuvipanda: Probably labs first. Maybe only ever labs. But would be nice to have the option to go do prod (hypothetically) [20:55:36] meeple27: :) true [20:55:36] so you agree that none would be problematic? [20:55:37] meeple27: nope :) [20:55:37] meeple27: well, as long as there’s a debian package for it. [20:55:37] meeple27: someone once wanted to run it on tcl 8.8 and there was no packages... [20:55:37] makes sense. thanks. [22:04:56] ^demon|lunch: do you use the mediawiki/extensions.git repo to have a copy of everything on your laptop or do something else entirely? I was trying to us my copy of mediawiki/extensions.git to do "real work" today and found it to be practically impossible to change things to push to gerrit. [22:07:26] bd808: I use it [22:07:43] how do you find it impossible? [22:08:06] the git remotes are hardcoded to https [22:08:22] rather than being relative to the base repo [22:08:48] Do you just add a new gerrit remote to each extension? [22:08:57] I have origin set to https, and 'gerrit' set to ssh, and run "git submodule foreach "git review -s"" every so often [22:09:20] hmmm [22:10:31] I have my git-review set to use origin for the gerrit remote. It seems like when I have run git-review -s with that setup on an https clone it blows up. I'll try again though [22:10:50] oh, I have it use 'gerrit' for the gerrit remote [22:11:40] at my old college, they blocked ports over a thousand, so I had to use https for cloning/pulling, so that was 'origin', and 'gerrit' was for over ssh [22:13:25] confirmed that `git-review -s` with my config blows up [22:15:15] :/ [22:22:49] bd808: hmm, I always use separate origin and gerrit remotes. [22:22:59] because cloning from gerrit was way slower than cloning from github when I was in India [22:27:33] yuvipanda: hey while you're about... I've got a ganglia + puppet config question that you might be able to help with [22:27:41] This cluster -- https://ganglia.wikimedia.org/latest/?c=Logstash%20cluster%20eqiad [22:27:56] is showing 4 of the 6 hosts [22:28:17] and the two that aren't showing are ones where I set $ganglia_aggregator = true [22:28:51] https://github.com/wikimedia/operations-puppet/blob/production/manifests/site.pp#L1345-L1350 [22:29:05] I was wondering if that was a bad cut-n-paste thing to do [22:29:28] bd808: ah, I’ve never touched ganglia :| [22:29:33] so not sure at all. [22:29:38] * yuvipanda was involved in graphite only [22:29:50] bd808: mutante maybe? He’s been migrating things to ganglia_new everywhere, so probably knows more [22:29:57] * bd808 found something yuvipanda doesn't know! [22:30:11] *nod* I'll just email ops-l [23:35:45] <^demon|lunch> bd808: I also don't use ssh or git-review, so yes the extension repo works for me :p [23:55:17] bd808: yes, git-review is horrible [23:55:19] well [23:55:20] not horrible [23:55:22] but I don’t like it [23:55:26] I just use it for git review -s