Swarming: Creates Feature Flag to control its usage#5190
Swarming: Creates Feature Flag to control its usage#5190
Conversation
letitz
left a comment
There was a problem hiding this comment.
If it were me, I would put this flag in the yaml config at the root instead of in the database. Seems to me that even if we want to Turn It Off Now, we could do a rollout with a new config file / rollback to the previous version, instead of flipping a bit in the database (which leaves no audit trail, requires no review, and makes an instant global config change).
I defer to ClusterFuzz maintainers however on their method of choice for gating new functionality.
Hi, I think @javanlacerda would be the best person to provide details on the chosen approach. Your suggestion seems good to me. |
| self.assertEqual(feature_flags.FeatureFlags.TEST_FLOAT_FLAG.string_value, | ||
| 'checker') | ||
|
|
||
| def test_not_found_in_db(self): |
|
Controls pull
…On Sat, Mar 7, 2026, 4:14 AM André Nogueira Ribeiro < ***@***.***> wrote:
***@***.**** approved this pull request.
—
Reply to this email directly, view it on GitHub
<#5190 (review)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/B332O6IQA2ESHLEW4CHQDL34PNEUFAVCNFSM6AAAAACWHLJXW6VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZTSMBWGMYDGMZVGU>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
Continuing the swarming efforts, we add a new Feature Flag to control wether swarming tasks are allowed or not to be scheduled. so for this:
is_swarming_taskto check for the feature flagdevenvironment test proofThis feature flag has been present in the
devenv DB since last monday, but this changes were present indevunder the previous swarming state (see this PR), still the metrics seemed healthy from monday to wednesday , worth nothing that it has been set on false:So today i send them back again to
devto test this changes after the rollback and metrics continue to be healthy and no regressions seem to be introduced cause of this error. The only noticable error i saw in the logs is: