mozilla :: #rust-infra

 
27 Apr 2017
00:02frewsxcvi'm looking forward to people getting r+ privileges whenever this team becomes a thing
00:07aidanhsI find something very satisfying about two fixes for spurious failures being merged into master sequentially
00:08aidanhs(aside from the obvious of there being fewer spurious failures...assuming they both work!)
00:13aidanhsfelicitous seems like a fitting word, I like it
00:24aturonsimulacrum: i don't think it needs to involve the whole core team, but i'm also not personally sure what all the implications are
00:24aturonin general we should feel empowered to make changes to infra (with buy-in from stakeholders, all of which are part of the team)
00:25simulacrumaturon: It's not too urgent, but the free heroku version of perf.rlo will not be running the whole month next month
00:26aturonsimulacrum: do you think you could make a (very brief) writeup of the current state, what you'd like to move to, and what the implications are?
00:26simulacrumNo problem
00:26simulacrumAny particular place? Google doc?
00:26aturonsimulacrum: google doc sounds good for now, yeah
00:26aturonwe'll have to figure out what process we ultimately want for review, but let's keep it super lightweight for now
00:28aturonacrichto: would you mind documenting, somewhere, the various permissions for things like r+, committing to the repo/labeling, where the highfive assignments are -- all the stuff along those lines? either in the FAQ or directly in the forge?
00:28aturonacrichto: i bring this up in part because i want to clear up what frewsxcv re: permissions appropriate for everyone involved in infra work
00:28aturon(i believe you were handing out some perms before, but am not sure exactly what you did)
00:29frewsxcvi know you probably still want documentation for it, but here's where reviewers are specified
00:29frewsxcvhttps://github.com/alexcrichton/rust-central-station/blob/master/homu.toml.template
00:31aturonfrewsxcv: thanks
00:31aturonand yeah, want to gather this stuff in a persistent place
00:38aturonaidanhs: just saw your PR, awesome work!
00:51acrichtoaidanhs: lol my favorite is when a "fix" for a spurious failure bounces because it fails spuriously
00:51acrichto"well I guess that answers that question"
00:52simulacrumacrichto: Do you know if the ALLOW_PR=1 current builder produces sufficient artifacts to allow running performance tests against? I expect those aren't uploaded anywhere, but it'd be interesting to know if we already produce them..
00:52acrichtoaturon: sure I will try to document that in the FAQ
00:53acrichtosimulacrum: I don't believe so, no
00:53acrichtobuilders on PRs don't upload anything
00:53simulacrumSure, but do they build enough?
00:53simulacrumI assume yes
00:53acrichtoI think so yeah
00:54acrichtob/c they don't build less than the main ones
00:54acrichtothey just skip the deploy step
00:54simulacrumI was thinking time wise more so than deploy wise
01:09simulacrumaturon: Dropped the document into the folder, but: https://docs.google.com/document/d/1PkKPq2gLaiOUXGVMRS0gqhIxo-MVkZxLR9qozWOPHMI/edit
01:09simulacrumI don't really know
01:09simulacrumIs that what you were looking for?
01:28aturonsimulacrum: perfect; only thing missing is the $ price
01:33simulacrumaturon: Hmm, I'll look into that
01:37simulacrumaturon: Added pricing. It's somewhat vague, but $0-25/month, likely $7/month I think
01:39aturonsimulacrum: ok, so a non-issue :) thanks again! last step, send out an email to infra-team@rust-lang.org with a link, asking for feedback in the form of comments on the google doc?
01:39aturonsimulacrum: then by the meeting we should be in a position to make a decision
01:40simulacrumSounds good
01:45est31what happened to perf.rlo?
01:45est31https://rusty-dash.com/links
01:45est31first link in this link
01:45est31list in this list*
01:46simulacrumest31: So perf.rlo I believe is hosted in EC2, which is difficult for us to give out access to
01:46est31so its down due to a bug?
01:46simulacrumEffectively yes
01:47est31okay
01:47simulacrumBut deploying to it is also hard since I can't do it
01:47simulacrumI added a note to the doc
01:47est31good to know, thought it got discontinued with the move to abandon buildbot
01:48est31oh you were discussing it right now.
01:52est31simulacrum: afaik, crates.io is already hosted on heroku
01:52est31maybe they offer cheaper access to two machines instead of one?
01:54simulacrumI don't think so, but it's possible
01:54est31https://internals.rust-lang.org/t/ipv6-reachability-of-rust-websites/4379/3?u=est31
01:54simulacrumLooks like $7/month is fine though so probably not too big a concern
01:54est31here is the link that says crates.io is on heroku
01:55simulacrumA source! :)
02:36frewsxcvanyone around?
02:36frewsxcvhttps://travis-ci.org/rust-lang/rust/builds/226222068
02:36frewsxcvrollup builds don't have any logs for me
02:37frewsxcvbuild times look normal though
02:37frewsxcvjust waiting on windows
02:37frewsxcvi'm going to blame this on travis
02:39acrichtofrewsxcv: travis has some db maintenance right now I think
02:39acrichtowhich may explain that
03:09simulacrumacrichto: Can we make https://buildbot.rust-lang.org/homu/queue/rust either redirect to buildbot2 or move buildbot2 to buildbot? I just get a hanging page on buildbot.r-l.o which is.. unfortunate.
03:09simulacrumNot sure if you're the person to talk to about this
04:14acrichtosimulacrum: we could yeah
04:14acrichtoI'd need to set up ssl certs
04:14acrichtowhich just takes effort
04:14acrichtonot as easy as a switch flip
04:19acrichtosimulacrum: if you're willing to give it a stab I could walk you through the steps
04:19acrichtoin theory it's just a PR to rust-central-station
04:20acrichtoand then I'd configure the DNS
06:00acrichtoaidanhs: I think this'll massively speed up android https://github.com/rust-lang/rust/pull/41575
08:32aidanhsa bunch of PRs have just failed on appveyor, which is ominous
08:35aidanhshmm, possible appveyor network problems
08:36aidanhs"Inner Exception: The remote name could not be resolved: 's3.amazonaws.com'"
08:47aidanhsindeed, https://appveyor.statuspage.io/
08:47aidanhs"Investigating - We are receiving reports from customers getting "remote name cannot be resolved" errors in their builds."
08:56aidanhsapparently fixed, let's see...
09:08aidanhsseems ok
12:42simulacrumacrichto: Yeah, sure, I'd be happy to try making buildbot.rust-lang.org work once more. Grepping for buildbot in central station only showed me one instance, though. I'll ping you later to discuss steps, need to run soon
13:28carols10centsi think this is travis, not us, but this osx worker just didn't do anything at all...https://travis-ci.org/rust-lang/rust/jobs/225569448 I logged it in the PR tracking spreadsheet
13:31carols10centsand i restarted the build so now it looks like everything was fine
13:31carols10centsthere was just literally nothing in the box where there's usually logs
13:46aidanhscarols10cents: sometimes that happens with osx builds and you only end up being able to see the logs a few hours after the build finished
13:47aidanhsin fact I saw it yesterday for a non-osx build
13:47carols10centsaidanhs: it was marked as failed tho
13:49aidanhsyeah, that sounds like the one I saw yesterday - powerpc linux had failed on a rollup but frewsxcv couldn't see the logs to find out why (I checked and was also unable to see the logs). I came to it a few hours later and they were there
13:51aidanhswhere 'a few' == 6, looking at the irc logs
13:53carols10centsstrange. oh well. i restarted it anyway
13:53aidanhsthough I didn't see the issue 'in action', so it could be completely different
13:53aidanhsprobably best - given it's a PR build, there's nothing for OSX to fail on anyway!
14:06acrichtosimulacrum: ok cool just lemme know
14:07acrichtoaidanhs: thanks for retrying those prs
14:07acrichtoand finding that status page!
 
27 Apr 2017
   
Last message: 53 minutes and 46 seconds ago