New issue
Advanced search Search tips

Issue 833191 link

Starred by 2 users

Issue metadata

Status: Duplicate
Merged: issue 832747
Owner: ----
Closed: Apr 2018
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 1
Type: Bug



Sign in to add a comment

chromiumos-sdk-tryjob hitting execution timeout at 12hrs

Project Member Reported by cmt...@chromium.org, Apr 15 2018

Issue description

I'm trying to figure out why the chromiumos-sdk-tryjobs are consistently failing in the InitSDK stage (see https://cros-goldeneye.corp.google.com/chromeos/legoland/builderHistory?buildConfig=chromiumos-sdk-tryjob)

I can pull up the overall build log, e.g. https://cros-goldeneye.corp.google.com/chromeos/healthmonitoring/buildDetails?buildbucketId=8949229384581453168

But when I click on the 'stdout' link, under the failed InitSDK stage, I get the log dog page with "Loading Streams" at the bottom.  It loads the streams for a couple of minutes and then it just stops.  The message across the bottom of the screen is:

"Loading streams (has the build crashed?)"

It has NOT loaded the entire log, I cannot set the error messages, and this is particularly useless.  Is there ANY way I can get the entire log, so that I can try to figure out why the builds are failing?
 

Comment 1 by cmt...@chromium.org, Apr 15 2018

typo: I cannot *see* the error messages (the loaded streams haven't gotten that far!)


Comment 2 by d...@chromium.org, Apr 16 2018

Cc: akes...@chromium.org dgarr...@chromium.org
Components: -Infra>Platform>LogDog Infra>Client>ChromeOS
Summary: chromiumos-sdk-tryjob hitting execution timeout at 12hrs (was: Luci/logdog fails in the middle of "streaming logs" for very large logs...)
If you click the link at the top, you can go to the actual build:

https://ci.chromium.org/p/chromeos/builds/b8949229384581453168#

It looks like the build was terminated. Clicking on the Swarming task, it says that the build timed out: https://chrome-swarming.appspot.com/task?id=3cdf34a4fbb69810&refresh=10&show_raw=1&wide_logs=true

LogDog doesn't know when a log stream has finished until the stream has declared that it is finished. If the stream never makes that declaration, LogDog optimistically continues to try and load new data until either that termination message arrives of the stream times out (typically 24hrs, forget if CrOS has a longer timeout), at which point it implicitly adds a termination.

None of that really matters, though, b/c what you're seeing is accurate: the last log in that stream is the last log in the build. The build was killed before it could emit additional logs. The log stream is complete.

Looking deeper into the Swarming task page, the build timed out because it ran for 12hrs and is configured with an execution timeout of 12hrs. Nice and straightforward :)

I'd check with someone in the CrOS team to determine whether or not 12hrs is appropriate for this build and if the timeout is due to something hanging when it shouldn't have, or just an excessively long number of operations. I'll CC some people now and re-purpose this bug for that investigation.

Comment 3 by cmt...@chromium.org, Apr 16 2018

Cc: manojgupta@chromium.org
Mergedinto: 832747
Status: Duplicate (was: Untriaged)

Sign in to add a comment