835089 - buildjson/builds-running.js incorrectly listing jobs for mozilla-b2g18_v1_0_0

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[Justin Wood (:Callek)]

So having triggered a handful of jobs on b2g18_v1_0_0 today, via "Rebuild" I don't see any jobs as pending/running when I look at it. Though it does show me the completed jobs.

I don't know if one of the regex's in here are causing us pain, or if its a releng bug. For the moment, since self-serve shows the stuff fine I'm going to assume TBPL bug.

I'll leave it to John if this is a weekend blocker, but I suspect not.

Comment 1

[Ed Morley [:emorley]]

Running builds on TBPL are client-side generated using https://secure.pub.build.mozilla.org/builddata/buildjson/builds-running.js 

Inspecting that file shows b2g18_v1_0_0 buildernames matched against the b2g18 tree.

-> Releng 

(Note: tbpl uses the same regexes for everything, so you'd either get no jobs shown, or all. The only issue that has been known to happen in the past is recent *completed* jobs for all trees not showing, due to TBPL's data import cron getting stuck. However the error handling for that has been recently improved, so even that shouldn't occur from now on. Best bet in the case of TBPL issues is to refer to the json file provided by releng to see whether it's TBPL's fault or a buildbot/releng script issue :-)).

Comment 2

[Justin Wood (:Callek)]

Attachment: [buildapi] v1

As far as I can tell this is the code that needs to change...

This should also fix Bug 751066, though I went with this approach in part because of how that bug played out (needing to be backed out).

To help trace callstack this is what I saw:

Running with fmt=json: http://mxr.mozilla.org/build/source/buildapi/buildapi/controllers/running.py

Index of results page: http://mxr.mozilla.org/build/source/buildapi/buildapi/controllers/results.py#25

Getting running builds list: http://mxr.mozilla.org/build/source/buildapi/buildapi/controllers/results.py#46

Figure out branch name based on results from buildbotDB: http://mxr.mozilla.org/build/source/buildapi/buildapi/model/query.py#116

We then see the "mozilla-b2g18" info before we see the "mozilla-b2g18_v1_0_0" and because we do startswith we leave the loop early and say we're "mozilla-b2g18": http://mxr.mozilla.org/build/source/buildapi/buildapi/model/query.py#34

Comment 3

[Justin Wood (:Callek)]

Attachment: [buildapi] v1.1

...I wonder what language I coded that in, since .length() is not actually the python syntax. This should do it

Comment 4

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

Comment on attachment 706824 [details] [diff] [review]
[buildapi] v1.1

This would cause regressions, where sourcestamps with branches like accessibility-linux-talos would be left alone rather than returning accessibility.

Comment 5

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

Attachment: comparison, using attachment 706824

See cruncher:~nthomas/bug835089 for how this was generated. 'grep -v OK' for the failures. Mostly talos, some release branch.

Comment 6

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

Attachment: comparison, attachment 706824 + fix

This has the changes you want for mozilla-b2g18_v1_0_0 and try-comm-central. I'm not sure what the story with those release branches is, they may be obsolete now but we'd have to look into it.

This requires this additional patch:

diff --git a/buildapi/model/query.py b/buildapi/model/query.py
--- a/buildapi/model/query.py
+++ b/buildapi/model/query.py
@@ -9,8 +9,9 @@ import math, re, time
 def GetAllBranches():
     ss = meta.scheduler_db_meta.tables['sourcestamps']
     q = select([ss.c.branch]).distinct()
     q = q.where(not_(ss.c.branch.like("%unittest")))
+    q = q.where(not_(ss.c.branch.like("%talos")))
     results = q.execute()
 
     # exclude defunct branches
     exclusions = ('releases/mozilla-1.9.3',

Comment 7

[Justin Wood (:Callek)]

Comment on attachment 706943 [details]
comparison, using attachment 706824 [details] [diff] [review]

I say we go with it, and look into the releasy stuff, we don't exactly support self serve for release jobs anyway [yet].

Comment 8

[Justin Wood (:Callek)]

Attachment: [buildapi] v2

Do it, and exclude talos

Comment 9

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

Comment on attachment 707421 [details] [diff] [review]
[buildapi] v2

http://hg.mozilla.org/build/buildapi/rev/4651e82bc894

Comment 10

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

Got autodeployed without issue, and looks good in production.