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

Issue 728665 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Last visit > 30 days ago
Closed: Jun 2017
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 0
Type: Bug



Sign in to add a comment

x86-alex-no-vmtest-pre-cq pre-cq jobs still run, after the config is removed from configuration.

Reported by jrbarnette@chromium.org, Jun 1 2017

Issue description

The x86-alex-no-vmtest-pre-cq builder is failing, because we've stopped
building Chrome for x86-alex.  The failures are blocking CLs, because
this is one of the default required builders for the pre-CQ.

The builder needs to be replaced with a pre-cq builder for a more
recent (and supported) board.

 
Owner: dgarr...@chromium.org
Status: Started (was: Assigned)
Where do you see this happening? I don't see it being listed as a PreCQ builder any more.
Owner: jrbarnette@chromium.org
A CL link would be enough.
Cc: pho...@chromium.org
Owner: dgarr...@chromium.org
> Where do you see this happening? I don't see it being listed as a PreCQ builder any more.

The system has been sending alert mail messages for nearly a week.
The most recent one is timestamped today at 10:52 AM.  That message
mentions this link:
    https://luci-milo.appspot.com/buildbot/chromiumos.tryserver/no_vmtest_pre_cq/52875

Looking in cbuildbot/constants.py, I see this:
# The default list of pre-cq configs to use.
PRE_CQ_DEFAULT_CONFIGS = [
    # TODO(ihf): Make caroline or newer vmtest once  crbug.com/708715  is fixed.
    'caroline-no-vmtest-pre-cq',      # skylake      kernel 3.18
    'daisy_spring-no-vmtest-pre-cq',  # arm          kernel 3.8
    'lumpy-no-vmtest-pre-cq',         # sandybridge  kernel 3.8
    'rambi-pre-cq',                   # baytrail     kernel 4.4       vmtest
    'samus-no-vmtest-pre-cq',         # broadwell    kernel 3.14
    'whirlwind-no-vmtest-pre-cq',     # brillo
]

I can't explain why the pre-CQ is still using x86-alex...  Perhaps we
need a pin bump?

No pin bump, but there there are other places that can refer to it (commit queue ini files). It's also possible for people to manually schedule schedule those builds, and that seems like the most likely candidate.

Bernie? Are we ready to fully remove mario/alex, etc from the chromeos_config.py?
Owner: bhthompson@chromium.org
Cc: akes...@chromium.org
Summary: x86-alex-no-vmtest-pre-cq pre-cq jobs still run, after the config is removed from configuration. (was: Remove x86-alex-no-vmtest-pre-cq from the pre-cq)
Aviv explained this.

We decide which PreCQ tests to run for a given CL once, and store that data in CIDB. We keep using that stored data until the CL is submitted, or a new revision is uploaded.

I'll send out a PSA about this, and consider it closed.
I sent out:

PSA: x86-alex-no-vmtest-pre-cq PreCQ failures

TL;DR
Rebase your CL if you get PreCQ failures for x86-alex.

Details

We recently removed a number of EOL'd boards from various parts of the build system. Because prebuilt binaries for Chrome are no longer available, they are not expected to build for PreCQ build configs. We removed these boards from the PreCQ configs, BUT CLs uploaded before we removed them still try to use the old list of builders.

If you upload a revised (even just rebased) version of the same CL, it will re-select the builders to test against, get an updated list, and be tested normally.

Sorry for the inconvenience.
Status: Fixed (was: Started)

Sign in to add a comment