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

Issue 687643 link

Starred by 1 user

Issue metadata

Status: Archived
Owner: ----
Closed: Jun 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 1
Type: Bug

Blocked on:
issue 279618



Sign in to add a comment

cyan-chrome-pfq failed HWtest due to timeout

Project Member Reported by x...@chromium.org, Feb 1 2017

Issue description

See the latest build:
https://uberchromegw.corp.google.com/i/chromeos/builders/cyan-chrome-pfq/builds/707

Selected Log info:
Autotest instance: cautotest
02-01-2017 [03:22:22] Submitted create_suite_job rpc
02-01-2017 [03:22:33] Created suite job: http://cautotest.corp.google.com/afe/#tab_id=view_job&object_id=99175474
@@@STEP_LINK@Link to suite@http://cautotest.corp.google.com/afe/#tab_id=view_job&object_id=99175474@@@
--create_and_return was specified, terminating now.
Will return from run_suite with status: OK
03:22:35: INFO: RunCommand: /b/cbuild/internal_master/chromite/third_party/swarming.client/swarming.py run --swarming chromeos-proxy.appspot.com --task-summary-json /tmp/cbuildbot-tmp2uXgUr/tmpNywxZW/temp_summary.json --raw-cmd --task-name cyan-chrome-pfq/R58-9240.0.0-rc1-arc-bvt-cq --dimension os Ubuntu-14.04 --dimension pool default --print-status-updates --timeout 18600 --io-timeout 18600 --hard-timeout 18600 --expiration 1200 '--tags=priority:PFQ' '--tags=suite:arc-bvt-cq' '--tags=build:cyan-chrome-pfq/R58-9240.0.0-rc1' '--tags=task_name:cyan-chrome-pfq/R58-9240.0.0-rc1-arc-bvt-cq' '--tags=board:cyan' -- /usr/local/autotest/site_utils/run_suite.py --build cyan-chrome-pfq/R58-9240.0.0-rc1 --board cyan --suite_name arc-bvt-cq --pool bvt --num 3 --file_bugs True --priority PFQ --timeout_mins 250 --retry True --minimum_duts 3 --suite_min_duts 3 --offload_failures_only False -m 99175474

@@@STEP_FAILURE@@@
04:10:08: ERROR: Timeout occurred- waited 15614 seconds, failing. Timeout reason: This build has reached the timeout deadline set by the master. Either this stage or a previous one took too long (see stage timing historical summary in ReportStage) or the build failed to start on time.
04:10:09: INFO: Running cidb query on pid 32470, repr(query) starts with <sqlalchemy.sql.expression.Insert object at 0x4f834d0>




 

Comment 1 by x...@chromium.org, Feb 1 2017

Cc: steve...@chromium.org

Comment 2 by x...@chromium.org, Feb 1 2017

Cc: uekawa@chromium.org semenzato@chromium.org
Anyone seen on this timeout error?
So, this is a "master timeout" of 15,000 seconds or about 4h 20m.  For such failures it's a good idea to look at the build times on viceroy:

https://viceroy.corp.google.com/chromeos/build_details?build_config=cyan-chrome-pfq&build_number=707&experimental=False&duration=1d&refresh=-1

and compare to earlier builds, to see what stands out.

Comment 4 by x...@chromium.org, Feb 1 2017

semenzato@: Thanks for the suggestion! Compared with earlier builds, seems the "HWTest[sanity]" Stage takes more than half an hour than usual (though it still passed at last). 
Labels: current-issue
The graph of earlier builds makes me worry a bit.  Many of the green ones are not much shorter than the ones that timed out.

Comment 6 by x...@chromium.org, Feb 1 2017

Yes, especially cyan-chrome-pfq/692 which was passed but has duration of 15896 seconds, while the current one cyan-chrome-pfq/707 timed out at 15993 seconds. How do we determine the timeout number? 
I am not sure where those timeouts are set, or even if the time shown in that dashboard is the same as the time used to device when timing out.  You would have to ask, or look at the code.

Comment 8 by x...@chromium.org, Feb 1 2017

Cc: -akes...@chromium.org
Owner: akes...@chromium.org
Since it's infra related issue, assign to Infra deputy for further investigation or find an better owner. 

Comment 9 by x...@chromium.org, Feb 1 2017

A new Chrome PFQ was kicked off. Monitoring...
Cc: nya@chromium.org hidehiko@chromium.org
+hidehiko, nya fyi

https://viceroy.corp.google.com/chromeos/build_details?build_config=cyan-chrome-pfq&amp;build_number=702

build took 1h51m ?
and arc-bvt-cq test only had 48 minutes remaining. 
(but that's a pretty normal condition, I wonder why this failed in particular)


Uekawa, can you point to the source that sets the timeout for arc-bvt-cq?

My impression is that these jobs are killed by the master timeout, not the step-specific timeout.

Thanks!

Comment 12 by nya@chromium.org, Feb 2 2017

Even in passed cases, BuildPackage stage always took ~2h to finish, and 90% of time it is building chromeos-chrome.
Building chrome should be done in <30m. I guess we do not use goma? Why?

Comment 13 by nya@chromium.org, Feb 2 2017

FYI, I found Issue 279618 about enabling goma in Chrome OS build.

Comment 14 by x...@chromium.org, Feb 3 2017

Update: Recent builds (~10 runs) passed without timeout error. 

However, looking at the build time dashboard on viceroy: https://viceroy.corp.google.com/chromeos/build_details?build_config=cyan-chrome-pfq&build_number=707&experimental=False&duration=1d&refresh=-1. It doesn't seem like the green ones are much shorter than the ones that timed out. And this timeout error also seems happen once in a while (see https://uberchromegw.corp.google.com/i/chromeos/builders/cyan-chrome-pfq?numbuilds=100, e.g., build#707, build#702, build#682). 

Enabling goma as in Issue 279618 will definitely help decrease this type of error in the future.  
See also  issue 611139 .

I just did a random audit, and with ARC now part of the build and test stages, we are cutting things pretty close to the master timeout in general.

We need to prioritize either making goma work (issue 279618) or increase the master timeout. 

Cc: akes...@chromium.org
Owner: ----
We did a mass wipe of build slaves, which meant their next build needed to repo sync from scratch. Could have been a transient timeout.

Comment 17 by aut...@google.com, Feb 9 2017

Labels: -current-issue
Status: Unconfirmed (was: Untriaged)
Blockedon: 279618
Status: Archived (was: Unconfirmed)

Sign in to add a comment