754003 - Retrigerring a pending job fails

Status: RESOLVED DUPLICATE of bug 755191

(Release Engineering :: General, defect)

Description

[(no longer active)]

This seems to happen consistently, the error message that you'd get looks something like "Rebuild request for Win XPCShellTest debug failed. (Build 11302641 not found on branch oak)".

Comment 1

[Nick Thomas [:nthomas] (UTC+12)]

Should it fail gracefully, or are you trying to schedule several jobs without having to wait for the first one to actually start ?

Comment 2

[Chris AtLee [:catlee]]

Maybe tbpl is sending the wrong request? It should be POSTing to 
/self-serve/{branch}/request with request_id as a post parameter.

That error message makes me think it's hitting /self-serve/{branch}/build instead.

Comment 3

[Phil Ringnalda (:philor)]

Oh so many many separate bugs:

tbpl bug: learn which branches have coalescing disabled (that would be try, and try, and there's try), and don't display the retrigger UI for pending jobs on any other branches, because that's pointless - yay, I triggered 5 runs while it was pending, which will get me exactly 1 when they are coalesced

self-serve UI bug: do display UI to retrigger pending jobs on try, where it has a huge point, since people quite often push at 17:00, see that they got builds by 23:00 but have tests still pending, and want to retrigger tests because they know they need n rounds of talos to average out, or m rounds of reftest to be sure they killed a randomorange

tbpl bug: rather than using whatever "this.runID.replace(/^(pending|running)-/, '')" from http://hg.mozilla.org/users/mstange_themasta.com/tinderboxpushlog/file/1a3eeafc91d6/js/MachineResult.js#l121 is for pending builds, look up the request_id in the json, because that runID is not the request_id, and you can't use it as either a request_id or as the build_id which does not yet exist

So I think what you wanted to file was a tbpl bug, "the weird way we guess at things to use as request_id for rebuilds does not work for pending jobs," even though that's not going to do you any good with oak.

Comment 4

[(no longer active)]

Now I'm completely lost on what needs to happen next!

Comment 5

[Phil Ringnalda (:philor)]

What needs to happen next, to get what you really wanted to say in comment 0, "I want to push to Oak, have the builds finish and schedule tests, and then while Windows xpcshell is still pending because there aren't enough Windows test slaves, I want to retrigger it and wind up with two xpcshell runs on that rev"? Bug 690672 needs to happen next: it doesn't matter what happens to tbpl's UI, what happens to self-serve's UI, or what number tbpl passes to what function of self-serve's API, as long as self-serve jobs get coalesced, you will only get one xpcshell run.

But I took pity on you, and filed bug 755191, and filed bug 755192, and filed bug 755193, even though none of them will make a jot of difference to your desired task until bug 690672 gets fixed.

Randomly choosing from among those, I'm going to call this a dupe of bug 755191, since that was the visible part you reported, even though it wasn't the biggest or the first boulder standing between you and your goal.