mozilla :: #taskcluster

14 Mar 2017
02:54akianyone else notice that taskGroupInspector has tasks blink on and off?
02:54* aki tries a different browser
02:55akihm, maybe it's a Nightly bug
03:05akinope, happening on chrome. not clear if it's when i have multiple tabs with different taskGroups open
03:08akinope, happening with a single tab in chrome
03:08* aki files bug
03:13firebotBug 1347029 NEW, in-progress Task Group Inspector graph "breathes"
13:10dustinjonasfj: no, I filed them like 24h earlier :)
15:03garndtpmoore: are you able to make the team meeting now? I know that daylight savings has screwed up the calendar for the next two weeks
15:04jhfordis pi now 14.3?
15:04jhfordinteresting ;)
15:10dustinjhford: no pie for you :)
15:21dustinjonasfj: scratching my head over - that file is present
15:21dustinand the /app dir gets MKDIR'd in the previous command
15:26dustinit worked for me locally -- must be a docker version thing
15:28bstackjhford: quick before you log off for the night: I want to steal a bit of tc-lib-logs and put it in tc-lib-monitor for structured logging of specific events. So rather than taking a debug-compatibility approach, it will be for specific logs like an auth audit log.
15:28bstackwe could work on lib-logs integration later if/when we want to make all of our logs structured
15:28bstackdoes that sound reasonable to you?
15:28jhfordbstack: is there a reason to not use lib-log in lib-monitor?
15:29jhfordrather than reimplementing it
15:29bstackjust because it the approach it takes is getting everything set up for debug-compat. I would import it then just use it for bunyan
15:29bstacksince I don't want any of the other stuff it does
15:29bstackbasically I just want to put bunyan in tc-lib-monitor
15:30jhfordso lib-log isn't all about debug compat
15:30jhfordthat's only there for transitioning things
15:30jhfordit's about bunyan + a DEBUG environment variable
15:30jhfordwell, equivalen tot it
15:30bstackyeah, I just don't want the debug environment stuff either
15:30bstackreally I don't even need bunyan
15:31bstackI guess I could just have tc-lib-monitor accept json blobs
15:31bstackthis loggings stuff wouldn't involve levels
15:31bstackperhaps that's the right approach here
16:02grenadecheers for the repoflag dustin!
20:21mjfIm looking for some help with the One Click Loaner link under Job Details on try. When I click on that link, I get a page that doesnt seem to do much. Task ID field with a pre-filled id, submit button (that doesnt seem to do anything), and I cant click on One-Click Loaner or Edit and Create Loaner Task. I am logged into with my moz ldap.
20:22KWiersosame here
20:22KWiersono errors get logged when I load the page, all network requests are successful 200 statuses, clicking "submit" doesn't throw an error, nor does it send any network requests
20:23KWiersowould be nice if there were tooltips showing reasons why the "one click loaner" and "edit and create loaner task" buttons were disabled
20:30mjfI bit more data - for a successful job on linux, the One-Click Loaner button seems to be enabled (havent clicked it yet). The unresponsive page was from a windows build that fails.
20:31garndtah, so in this case it's because Windows and OS X do not support one click loaners
20:32mjfOf course, it is exactly the windows build that Im trying to troubleshoot with the one click loaner. :-)
20:33mjfI can click on a OSX build and the button is enabled.
20:34mjfAh, and then for either OSX or linux I get a Insufficient Scopes Error if I try to click through.
20:35garndtah sorry, we do not support one click loaners for OS X test tasks...the build tasks are not built on macs but rather cross compiled on linux
20:36garndtI wasn't clear, I'm sorry
20:38jmaherbstack: I am trying to backfill android debug r26 on autoland and it doesn't seem to be working
20:38mjfgarndt: I suppose that it really doesnt matter since the windows part is the one I really need, but isnt supported, but I am currious about how to correct the Insufficient Scopes Error for linux one click loaners.
20:38jmaherbstack: this is the job I am trying to backfill:
20:38jmaheroh, red action tasks
20:39jmaherhmm |The taskgraph command does not accept the arguments: --treeherder-url|
20:39bstackperhaps somebody changed something
20:39bstackactually, how old is that branch?
20:39bstackoh, autoland
20:40jmahermarch 8th
20:40garndtmjf: depends on what the scope error was, mind opening up a bug in Taskcluster::Service Request with the details of the error?
20:40jmaherso ~1 week old, not 4 weeks old
20:40* bstack updates firefox checkout
20:40bstackone sec, I'll look
20:40bstackI wonder if treeherder was just deployed
20:41rwoodbstack: yep
20:41jmaherbstack: yeah, not long ago
20:41jmaherabout 1 hour ago I had logged out/in on treeherder
20:43bstackthe treeherder-url was just to get it working in staging too. I'm sad that it broke things in production
20:43bstackI'm checking up on bugs real quick, one moment
20:43jmaherok, I will add new jobs to fix things then
20:45jmaherbstack: no need to feel bad, perf alerts were broken ~1 hour ago :)
20:46mjfgarndt: Bug 1347282
20:46firebot NEW, One-Click loaner gives me "Insufficient Scopes Error!"
20:46garndtthank you!
20:46bstackah, the treeherder change got landed, but not the gecko change
20:46firebotBug 1333604 FIXED, make action tasks that interact with treeherder able to accept a treeherder api url
20:46bstackweird, it's r+ in bugzilla but r? in reviewboard
20:47jmaherbstack: maybe the mozreview wasn't used to review it?
20:47jmaherI had that happen once
20:47bstackquickest thing here is to back out the treeherder change
20:47bstackI assume that the gecko code has bitrotted
20:47bstackbut if not, landing that will make this work in autoland
20:48bastienHello guys
20:49bastienI'm using taskcluster email notifications, to send some automated reports
20:49bastienIs there any way to get a "fullscreen" email template instead of the centered column one ?
20:49dustinI think there's only the one template
20:50* dustin ambivalent about giving the "it's in this repo if you want to fix that up" answer :)
20:50bastienthat's what i thought :p
20:50bstackbastien: we've considered having multiple templates available. it shouldn't be a hard change to make.
20:50bstackit would be ""
20:51bastienFYI i'm sending some mercurial error outputs, and the long multi lines kinda break the template
20:52bastienbstack: should i add an optional argument to the email endpoint to specify a template name ?
20:52bastienOf course it would use the "simple" one by default
20:53bstackjmaher: I'm talking in treeherder to see how to get that backed out
20:53jmaherbstack: ok, thanks- I did 'add new jobs' for the missing revisions and got it working
20:53bstackbastien: yeah, that would work
20:53bstackare you hitting the api or going through the task definition to send the emails?
20:54bastienhitting the api, through taskclusterProxy from a hook
20:55bstackthen yeah, should be quite easy. just make it configurable and make the api endpoint accept an argument for it
20:55bstackand then we can just generate a new client and it should all work
20:57bastienok cool
20:57bastieni'll make a PR on taskcluster-notify and ping you
20:57bstack++ :)
20:57bstackthanks a lot
20:57bstackI've always wanted that
21:09mcotejmaher, bstack: seems camd r+ed it in Bugzilla
21:09mcotecamd: my own report doesn't know how to use mozreview
21:09mcotefor shame
21:09bstackah yeah. I never asked for it to be landed though. that was my bad
21:10bstackI was trying to clear out a backlog of PRs a few days ago and didn't look at that hard enough
21:10bstackshould be fixed now. send thanks to wlach and KWierso :)
21:10bstacktotally my fault
21:10bstackbut all is well now
21:10mcotecamd landed it for you I guess?
21:11* KWierso flees
21:11bstacknah, I never asked for the m-c bits to be landed and when I looked again later I figured I had already and asked for the treeherder bits to be landed
21:12bstackpebkac on my side
21:12mcoteyeah multiple patches going to different repos in one bug is tricky :)
21:14KWiersotwo month old PR had plenty of time for the other side of things to land! :P
21:48* camd feels his ears buzzing...
21:50KWiersoI mean yes, yes it was all camd's fault
21:58bstackjonasfj: did you see ?
21:58bstackit's pretty much your PR but different :p
21:59jonasfjbstack: lol
21:59bstackso now we just get to pick whichever one we think handles it more correctly and click merge :p
21:59jonasfjyes :)
21:59bstackmine is a more annoying way of handling it
21:59bstackalthough potentially less surprising?
21:59jonasfjoverhead is probably the same..
21:59jonasfjwell, technically we're always allowed to return less than 1k
21:59bstackoh yeah, this library is supposed to be "fast"
22:00jonasfjbstack: your is more correct
22:00jonasfjthe Math.min solution belongs in azure-entities
22:00bstackah yeah, that's a good point
22:01bstackmaybe azure-entities can even have a nice debug line that says "silently changing foo to 1000" ?
22:03bstacktoday we fixed bugs!
22:03bstackfast-azure is in travis for publishing now
22:03bstackonce we get azure-entities, we can update queue and click resolve
22:03bstackoh wait
22:04bstackgithub rejected my push to master
22:04bstackbut not the tag of the new commit
22:04bstackthat's confusing
22:04jonasfjI'll let you do that, it just rejected my 1.0.1 tag too :)
22:05bstackok, so I'll make admins allowed to push directly?
22:05jonasfjthat's why?
22:05jonasfjI wish there was a PR review process for releases
22:06jonasfjbstack: yeah, admins can push I guess..
22:06bstackafaict there's no way with the travis publishing to require it both be on a branch _and_ have a tag
22:06bstackwhich is why they have to go direct to master
22:06bstackbut perhaps my knowledge is old/wrong
22:07jonasfjwe could bump version and push the tag later...
22:08jonasfjbstack: if crazy we could tweak travis to inject tag value in version field in package.json, that way we just hit the "tag release" button in github
22:08bstackah interesting
22:09jonasfjbut we're probably better off keeping it simple :)
22:10jonasfjI love that we just found a fix for another bug using sentry
22:10bstackSimple is boring!
22:11bstackYeah, sentry is pretty solid.
22:11dustin^^ this is why we can't have nice (software) things :)
22:16jonasfjlol, that's so true...
23:38bstackjonasfj: currently tc-auth can't give out sas creds for the auth azure account itself (which means it doesn't get backed up yet). I'd like to make it give out only read-only creds for it. The code changes are minor, but the tests currently only have one azure account for the azure tests.
23:38bstackTo test this correctly, it will need a second account. Shall I make a second jungle account?
23:39bstackimMemory will be hard to get to work with this
15 Mar 2017
No messages
Last message: 10 days and 16 hours ago