New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 774287 link

Starred by 2 users

Issue metadata

Status: WontFix
Owner:
Closed: Oct 2017
Cc:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 2
Type: Bug



Sign in to add a comment

chromite: config_dump.json doesn't match

Project Member Reported by laszio@chromium.org, Oct 12 2017

Issue description

https://uberchromegw.corp.google.com/i/chromeos/builders/caroline-release/builds/1118

chromite-0.0.2-r3428: FAIL: [chromite.cbuildbot.chromeos_config_unittest] ConfigDumpTest.testDump
chromite-0.0.2-r3428: Make sure the json & config are kept in sync
chromite-0.0.2-r3428: ----------------------------------------------------------------------
chromite-0.0.2-r3428: Traceback (most recent call last):
chromite-0.0.2-r3428:   File "/mnt/host/source/chromite/lib/timeout_util.py", line 191, in TimeoutWrapper
chromite-0.0.2-r3428:     return func(*args, **kwargs)
chromite-0.0.2-r3428:   File "/mnt/host/source/chromite/cbuildbot/chromeos_config_unittest.py", line 58, in testDump
chromite-0.0.2-r3428:     self.fail('config_dump.json does not match the '
chromite-0.0.2-r3428: AssertionError: config_dump.json does not match the defined configs. Run cbuildbot/chromeos_config_unittest --update

It started to happen in 10026. The only change between 10025 and 10026 is
crrev.com/9af4a9a23a7589ce4dec411604695c4572c75190

I'm not sure why so CC'ing ihf@ (the author).
 

Comment 1 by ihf@chromium.org, Oct 12 2017

Cc: dgarr...@chromium.org
Owner: ihf@chromium.org
Status: Started (was: Untriaged)
Will take a look. The tests passed for me, and they made it through the cq.
The chumped change must have done the wrong thing cherry-pick/merge wise.

Strictly speaking, this is possible even when going through the CQ because there are multiple entities writing to the file.
Oh... no, TOT is clean.
I believe this is a really rare unittest flake. I'm not sure why it happened through.

Comment 6 by ihf@chromium.org, Oct 13 2017

I built caroline locally and did not hit this.

Comment 7 by kinaba@chromium.org, Oct 13 2017

chrome-bot updated the config dump at the same cycle as Ilja's CL
https://chromium.googlesource.com/chromiumos/chromite/+/b740410d1b9ba8070f052699da5cac485d4d1185
and one more time in the next cycle
https://chromium.googlesource.com/chromiumos/chromite/+/31ef8f074175b067e95f12a50403dd9374c0c284

Perhaps the second one for R63-10027.0.0 adjusted and fixed it for ToT?

Comment 8 by ihf@chromium.org, Oct 13 2017

Cc: bhthompson@chromium.org
Labels: M-63
Good observation! This might be a race where the GE config change adding grunt (+VMTest) was overridden by my change to not run VMTest.

Bernie, do GE config changes go through the CQ or are they chumped? If the latter, then I think this can be explained and the redness is for a single build.

I tried to manually kick off a few more release builds.

Comment 9 by kinaba@chromium.org, Oct 13 2017

Labels: -Pri-0 Pri-2
(Release bots that grabbed R63-10027.0.0 is getting greener. Lowering the priority.)
Cc: nxia@chromium.org
+nxia 

The GE changes are 'chumped' much like how a PFQ run will 'chump' a Chrome uprev CL, they don't go through the formal CQ, but if the config updater builder lands something it should be consistent at the point in time the config updater synced, it is possible we have a race between the CQ and the config updater here.

The config updater builder can be observed at https://uberchromegw.corp.google.com/i/chromeos.infra/builders/config-updater

It is probably self correcting within an hour at least, I think the config updater will run again and correct any inconsistencies in the json.

Comment 11 by ihf@chromium.org, Oct 13 2017

Status: WontFix (was: Started)
Ok, I figure this is a rare event. Closing.

Sign in to add a comment