Closed Bug 1056787 Opened 10 years ago Closed 10 years ago

Skip every nth job for debug tests on mozilla-inbound

Categories

(Release Engineering :: General, defect)

x86_64
Linux
defect
Not set
normal

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: catlee, Assigned: catlee)

Details

Attachments

(3 files)

We'd like to run an experiment where we skip every nth job for debug tests on mozilla-inbound. The purpose of the experiment is to see by how much reducing the test load reduces overall infra load, while at the same time not putting undue burden on sheriffs if and when debug-specific failures arise.
Ben, what do you think about https://github.com/catlee/buildbotcustom/compare/master...skip_scheduling ?
Flags: needinfo?(bhearsum)
It looks reasonable to me. Almost too easy....
Flags: needinfo?(bhearsum)
Attachment #8476879 - Flags: review?(bhearsum)
Attachment #8476880 - Flags: review?(bhearsum)
Attachment #8476879 - Flags: review?(bhearsum) → review+
Comment on attachment 8476880 [details] [diff] [review]
enable skipping on mozilla-inbound

Review of attachment 8476880 [details] [diff] [review]:
-----------------------------------------------------------------

r=me if you add a comment about what can break down if you increase skipcount more than the global coalescing limit, per IRC.
Attachment #8476880 - Flags: review?(bhearsum) → review+
Attachment #8476879 - Flags: checked-in+
Attachment #8476880 - Flags: checked-in+
In production with reconfig on 2014-08-25 08:49 PT
Attached patch cleanupSplinter Review
removes MultiScheduler. Wasn't easy to remove the 'tests' count since that's used in some places to disable tests, so I added an 'assert tests == 1' instead.
Attachment #8480163 - Flags: review?(bhearsum)
Comment on attachment 8480163 [details] [diff] [review]
cleanup

Review of attachment 8480163 [details] [diff] [review]:
-----------------------------------------------------------------

(In reply to Chris AtLee [:catlee] from comment #7)
> Created attachment 8480163 [details] [diff] [review]
> cleanup
> 
> removes MultiScheduler. Wasn't easy to remove the 'tests' count since that's
> used in some places to disable tests, so I added an 'assert tests == 1'
> instead.

It seems that this is possible by setting tests = 0 or removing the slave platform.

> python -mthis | tail -n7 | head -n1
> There should be one-- and preferably only one --obvious way to do it.

Oh well...
Attachment #8480163 - Flags: review?(bhearsum) → review+
patch(es) in production :)
Status: NEW → RESOLVED
Closed: 10 years ago
Resolution: --- → FIXED
Component: General Automation → General
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: