mozilla :: #taskcluster

7 Aug 2017
07:54Tomcat|sheriffdutywcosta: ping
07:55Tomcat|sheriffdutyoh that might a little too early for wcosta :)
07:55Tomcat|sheriffdutyjhford: ping
07:55Tomcat|sheriffdutyjhford: did we change anything for osx this morning ?
07:56Tomcat|sheriffdutyor anyone else know ?
10:30wcostaTomcat|sheriffduty: i upgraded generic-worker version
10:31wcostaOps, no, patch didn't land actually
13:03teddustin: boy, i am having a hard time getting my upload-generated-sources task working
13:03tedi was trying to cargo-cult from the upload-symbols task, since it's similar in spirit
13:03tedbut that one is just quirky enough to make it hard
13:03tedi'm thinking now i might refactor upload-symbols into a more generic 'post-build' kind and then build off of that
13:06tedmaybe i'm overthinking this and i should just make my shit work and forget about upload-symbols for now
13:07jhfordi trust that you'll come up with something good :)
13:08tedi mean, conceptually we could probably take a bunch of these tasks that run after a nightly and smash them together
13:08tedbut it might not actually make things nicer
13:10jhfordi think having a common framework for these jobs would be worthwhile, if you used a declarative sytnax, you could share all the uploading/downloading logic
13:10jhfordi'm sure a lot of them are basically: run this bash locally and POST these files to these locations
13:10jhfordwhere TC-Artifacts is one possible location
13:13tedyeah, i dunno
13:13tedwe've got several nice abstractions here, but they expect things to work a certain way
13:14tedand then we've got a bunch of one-offs that leverage a little bit of other abstractions
13:15Callek...remind me, does taskclusters decision task trigger on pushloghtml stuff, or on generic pulse messages from hg.m.o directly (as in, does a broken pushlog delay jobs)
13:18jhfordahh, fair
13:22Callekdustin: jhford: ^ any knowledge on the answer to my question about 10 min ago?
13:30jhfordCallek: i can't say for sure, but we definitely have code for pushlog parsing and treeherder message listening in the relevant system. i'm not sure which one happens for real, but the pulse listening stuff is much newer
13:30Callekjhford: the impetus behind the question, is try pushlog is stuck for almost 2 hours now, with no new entries despite new pushes
13:31Callekand I was curious if we had already started jobs, or if we even *would* start jobs.
13:31jhfordi figured it would be related to a pushlog backup, based on the question
13:31jhfordCallek: you could try listening to the pulse messages using the site
13:32dustinCallek: mozilla-taskcluster polls pushlog (not htmlpushlog)
13:33jhforddustin: what do we use the pulse listener for then, out of curiousity
13:33Callekdustin: pushlog and pushloghtml are both busted right now
13:33Callek(well busted in that they are ~2 hours behind, on try)
13:33dustinjhford: in mozilla-tsakcluster? I don't know
13:33jhfordyah, there. ok!
13:33dustinoh, for retrigger and stuff like that
13:33jhfordyah, there. ok!
13:33jhfordirccloud is a little special today it seems
13:34jhfordahh, makes sense
13:34dustina lot of "we failed to send your message"
13:34jhfordthat's what I got
13:34dustinwhich it then, maybe?, sends
13:34jhforddid you see "yah, there. ok!" twice from me?
13:34jhfordyah, it shows as sent for me
13:35jhfordi wonder how bad of an idea it would be to have publish to pulse at the same time as sending data to pushlog, or if we've already tried that andt here's a reason not to
13:36dustinit does publish to pulse
13:36dustinwe have plans to write a hook that responds to those
13:36dustinand turn off mozilla-taskcluster
13:36dustinbut hooks doesn't respond to pulse messages yet :)
14:37tomprincedustin: Does hg publish to pulse, or does nozilla-taskcluster do the hg taskcluster do the publishing after polling the pushlog?
14:38tomprinceWell, the later definitely happens. I was wondering if the former also happens.
14:45dustinI don't know OTOMH
15:11aobreja|builddutyAs spoken last week,regarding Bug 1376807,I will need to change the password for OS X in secrets hiera and will need if possible as Amy suggested to coordinate with someone from taskcluster team to check the jobs in production after this change,I spoke with to do this in european morning time but last week there were lots of pending jobs and we
15:11aobreja|builddutycouldn't do this.Now Peter is on pto and the number of pendings is low,Peter told me that if we do the changes this week I can speak with wcosta or jhford for european morning.
15:11firebot NEW, Autologin is failing on the mac machines
15:12arraobreja|buildduty: I saw catlee asking that we not make any changes until after the beta ships this week
15:12arr(not about this specifically, but regarding a different change to OS X)
15:13aobreja|builddutyarr:ok,then we will find another time,thank you
15:13aobreja|builddutyarr: I will be on pto next week
15:14arraobreja|buildduty: would be good to talk to mtabara about timing around the beta release
15:15aobreja|builddutyif we can't do this change this week,maybe I can also just create the patch and someone else could do the changes
15:37mtabaraarr: by "different change to OSX" you meant
15:37firebotBug 1315977 NEW, Mac build machines doing l10n repacks need python 2.7.6 or newer
15:37arrmtabara: yep!
15:37mtabarayeah, talked to catlee about that on Thursday.
15:37mtabarawe're holding on from making any changes to any machines until we see 55.0 out of the door + beta 1 shipped.
15:38wcostaaobreja|buildduty: I am available to help whatever you need
15:38mtabarato avoid adding infra issues to an already pretty "fire"-ish period lately.
15:39mtabara55.0 is due tomorrow while 56.0b1 is due on Wednesday. let's aim for Thursday to coordinate with all actors involved
15:39mtabaraby then let's prep all the necessary pieces so that we can make this lethally sweet: fast and efficient :)
15:39aobreja|builddutywcosta:thank you,I will tell you when we can do the change,after we have more info
15:39mtabaraaobreja|buildduty: sounds good? ^
15:40mtabaralet me know if you need me to rope in more people or help anyhow. otherwise, let's aim for Thursday morning
15:40aobreja|builddutysounds good :)
15:41aobreja|builddutywcosta: Thursday morning 10-12 AM CEST is ok for you ,or is to early?
15:42aobreja|builddutymtabara:thank you :-)
15:43wcostaaobreja|buildduty: yep
15:44aobreja|builddutywcosta:great ^^
17:58gerard-majaxhey hello
17:58gerard-majaxit has been a while since I asked a silly question
17:58gerard-majaxartifacts are stored on S3, so we pay for them when ? Upon storage or upon each download ?
18:00tedif i use `kind-dependencies`, does that not automatically add a dependency to my output task?
18:01dustinit does not
18:01dustinit just orders the generation of kinds
18:04tomprincegerard-majax: Although the cost is ~$0.02/GB for storage and <$0.09/GB on data transfer out of amazon. So unless you have a lot of data, I wouldn&#39;t worry to much about it.
18:04teddustin: OK, interesting
18:05gerard-majaxtomprince, what&#39;s &quot;a lot&quot; ?
18:07teddustin: in the output of `./mach taskgraph tasks --json`, all the tasks have `&quot;dependencies&quot;: {}`
18:08dustinyeah, use ./mach taskgraph full
18:08Callekdustin: can you please add `project:releng:signing:format:mar_sha384` to try (moz-tree:level:1) Bug 1388118
18:08dustintasks is just the list of tasks, without dependencies calculated
18:08firebot NEW, Please add signing-related scopes to try
18:08tomprincegerard-majax: I don&#39;t know, since I&#39;m not responsible for taskcluster or its cost. But if you aren&#39;t generating several 100G, it probably isn&#39;t worth worrying about.
18:09garndtgerard-majax: what are you worried about? :)
18:09garndtor, I should rephrase, what crazy things are you about to do? I&#39;m kidding
18:10gerard-majaxI may be exposing some GBs artifacts
18:10teddustin: ahh
18:10teddustin: thanks
18:51dustinfubar: can you hook me up with AWS access to the releng account? I want to see how the VPCs are configured there
18:57fubardustin: I don&#39;t have admin there, I don&#39;t believe; got my acct from catlee
18:58dustincatlee: ^^?
18:58akihe&#39;s on pto all week
18:59dustinI&#39;ll fake it
8 Aug 2017
No messages
Last message: 14 days and 29 minutes ago