-
Notifications
You must be signed in to change notification settings - Fork 321
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[RayService] Address Recent Flakiness in RayService Zero Downtime Rollout Test #1979
[RayService] Address Recent Flakiness in RayService Zero Downtime Rollout Test #1979
Conversation
Signed-off-by: Yicheng-Lu-llll <luyc58576@gmail.com>
@@ -76,7 +76,7 @@ def __init__( | |||
self, | |||
custom_resource_object, | |||
rulesets: List[RuleSet] = [], | |||
timeout: int = 90, | |||
timeout: int = 180, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We now need to wait for the old RayCluster to be deleted if switching the RayCluster. The old RayCluster will only be deleted 60 seconds after it becomes inactive. Therefore, we need at least 90 + 60 = 150 seconds; the extra 30 seconds are for the overhead and safety.
I have triggered the CI test 5 times, and the RayService test has passed every time. The RayService Zero Downtime Rollout Test seems stable now. Note, the failed CI tests are RayCluster tests, which are unrelated to this PR. |
Why are these changes needed?
Recently, RayService zero downtime rollout test becomes flaky:
#1837 describes the root cause and introduces a hotfix.
My previous pr #1928 attempts to address the issue and remove the hotfix but as shown avoce, it actually fails to resolve them.
This PR introduces logic to ensure that, for zero-downtime rollout tests, during the period when the serve service has not been fully updated and the old RayCluster has not been deleted,
RayServiceUpdateCREvent
expects the query request to be processed by either the old or the new RayCluster.Related issue number
Checks