New issue
Advanced search Search tips

Issue 751271 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Closed: Oct 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug

Blocked on:
issue 600469
issue 770592



Sign in to add a comment

CQ tests leak Git daemons.

Project Member Reported by d...@chromium.org, Aug 1 2017

Issue description

CQ tests leak Git daemons, until the system can't bind new ports. This ultimately causes builders to fail.

We've paved over the problem for now by auto-rebooting buildslaves: https://chrome-internal-review.googlesource.com/c/422736

However, this is a bad bug and needs to be fixed, as similar leaks could happen on any system that runs these tests.

Example builds:
- https://luci-milo.appspot.com/buildbot/internal.infra/infra-internal-continuous-trusty-32/4250
- https://luci-milo.appspot.com/buildbot/internal.infra/infra-internal-continuous-trusty-32/4249
- https://luci-milo.appspot.com/buildbot/internal.infra/infra-internal-continuous-trusty-32/4248
- https://luci-milo.appspot.com/buildbot/internal.infra/infra-internal-continuous-trusty-32/4245
 
Blockedon: 600469
Labels: -Pri-1 Pri-2
Status: Available (was: Untriaged)
Yes, KI, but I punted fixing it in favor of killing all Rietveld CQ code out right.
I disagree this is Pri1 though, given your fix.

Comment 2 by d...@chromium.org, Aug 2 2017

I think P1 is appropriate because this test is run on developer machines as well, and having those lose gradually get all of their ports consumed seems bad to me.
Components: -Infra>CQ Infra>Platform>CQdaemon
Labels: -Pri-2 Pri-1
Owner: tandrii@chromium.org
Status: Assigned (was: Available)
I think this code has started failing more frequently than before.

I used to run CQ tests every time I flushed vim buffers, which means every ~10 seconds and never ran into this problem. I suspect new git version made this more likely.

The tree got closed again just now for the same reason. I'll monitor this task for another few days and then perhaps just delete these tests after verifying that major projects don't use Rietveld CQ.


Comment 4 by d...@chromium.org, Aug 2 2017

The tree closing is not surprising because the waterfall wasn't restarted to pick up my fix. I tasked someone with that and it looks like they didn't do it :/
Oh well, i'll restart it then: https://chrome-internal-review.googlesource.com/423369
Project Member

Comment 6 by bugdroid1@chromium.org, Aug 2 2017

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infradata/master-manager/+/c5d4dbc369bfda81acf63e5734f0245e87c5cfd5

commit c5d4dbc369bfda81acf63e5734f0245e87c5cfd5
Author: Andrii Shyshkalov <tandrii@google.com>
Date: Wed Aug 02 16:31:42 2017

Cc: tandrii@chromium.org
Labels: -Pri-1 Pri-2
Owner: ----
Status: Available (was: Assigned)
Major projects still use Rietveld CQ, though much less so. Downgrading to Pri2.

Comment 8 by efoo@chromium.org, Aug 31 2017

Components: Infra>Platform>CQ

Comment 9 by efoo@chromium.org, Aug 31 2017

Components: -Infra>Platform>CQdaemon
Owner: tandrii@chromium.org
Status: Started (was: Available)
Blockedon: 770592
Project Member

Comment 12 by bugdroid1@chromium.org, Oct 11 2017

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infra/infra_internal/+/2f06d5187315452ad90a14349c26fccd41630368

commit 2f06d5187315452ad90a14349c26fccd41630368
Author: Andrii Shyshkalov <tandrii@chromium.org>
Date: Wed Oct 11 23:48:56 2017

Tests are gone!
Status: Fixed (was: Started)

Sign in to add a comment