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

Issue 835292 link

Starred by 2 users

Issue metadata

Status: Fixed
Owner: ----
Closed: May 2018
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 1
Type: ----



Sign in to add a comment

Goma fails to start on WebKit Mac Builder due to system clock +5 min in the future

Project Member Reported by sheriff-...@appspot.gserviceaccount.com, Apr 20 2018

Issue description

Filed by sheriff-o-matic@appspot.gserviceaccount.com on behalf of jwd@google.com

d72e7ada-0c20-4e4d-860f-bd0380c4bccb

Builders failed on: 
- WebKit Mac Builder: 
  https://build.chromium.org/p/chromium.webkit/builders/WebKit%20Mac%20Builder

Exception steps exception preprocess_for_goma exception preprocess_for_goma.start_goma failed failure reason.

This has been happening for 2 hours now, but seems to have become more consistent in the last half hour.
 

Comment 1 by jwd@chromium.org, Apr 20 2018

 Issue 835293  has been merged into this issue.
Components: Infra>Goma
Labels: -Infra-Troopers Foundation-Troopers
Goma related issue, change the label to Foundation-Troopers.
Summary: Goma fails to start on WebKit Mac Builder (was: d72e7ada-0c20-4e4d-860f-bd0380c4bccb)
goma_compiler_proxy logs are not preserved (nor sent to Cloud Logging) in case goma fails to start, so it is hard to debug :-/ Trying to catch the bot in the middle of staring goma to grab logs via SSH.
There's a step "preprocess_for_goma.upload_goma_start_failed_logs" with something. But I can't find many details there: https://chromium-build-stats.appspot.com/compiler_proxy_log/2018/04/20/vm31-m1/compiler_proxy.vm31-m1.golo.chromium.org.chrome-bot.log.INFO.20180420-152409.68672.gz

This step itself also has errors from tsmon library:
Failed to retrieve access token: {
  "error" : "invalid_grant",
  "error_description" : "Invalid JWT: Token must be a short-lived token (60 minutes) and in a reasonable timeframe. Check your iat and exp values and use a clock with skew to account for clock differences between systems."
}

Which suggest issues with the clock. They probably affect goma client too. 

The clock on the machine in 5 min into the future. I think it is enough to screw up JWTs.

I ran "ntpdate -u time1.golo.chromium.org" to fix the time. Will look at the build to confirm it helped.
Components: Infra>Labs
Summary: Goma fails to start on WebKit Mac Builder due to system clock +5 min in the future (was: Goma fails to start on WebKit Mac Builder)
+ Labs, since I believe we are running ntpd automatically everywhere and this should not have happened.
Yeah, the problem was in the clock. Next build started goma successfully: https://ci.chromium.org/buildbot/chromium.webkit/WebKit%20Mac%20Builder/240394
The builder might not auto-reboot?
I heard that mac sometimes stops ntpd silently.  Bots with auto-reboot automatically restarts the job.  However, for bots without auto-reboot, ntpd is kept on dead and time would be unacceptable large difference for GoogleAPI.

If my memory is correct, we might have seen the same issue before, and we made the bot auto-reboot at that time.
Labels: -Foundation-Troopers
Status: Untriaged (was: Available)

Comment 10 by ukai@chromium.org, Apr 24 2018

already fixed?
or anything remain to do with this issue?

Comment 11 by ukai@chromium.org, May 17 2018

Status: Fixed (was: Untriaged)

Sign in to add a comment