-
Notifications
You must be signed in to change notification settings - Fork 332
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
[proposal] gang strategy enhancement #2244
Comments
I'm also experiencing this confusion, and I'm glad to see someone bringing it up. I would appreciate it if the author could take a look at this issue. |
I totally agree with this proposal. I've encountered similar issues where the gang policy behavior doesn't align with user expectations after a pod is destroyed and rebuilt. Keeping the gang semantics intact until the MinRequiredNumber is satisfied makes sense from a scheduling perspective. I support adding the user intention logic to address this. Thank you for raising this issue, and I look forward to seeing it resolved! |
I have encountered the same problem as the author. This proposal has been of great help to me. Thank you. |
@LY-today Thanks for your suggestions. Feel free to self-assign if you are willing to contribute or have been working on it! |
@saintube ps:I am very willing to submit the corresponding PR, and I will develop it after the communication is completed. |
What is your proposal:
Add user wishes comments to the logic in the gang policy to prevent the gang policy from directly invalidating after the pod is destroyed and rebuilt.
if gang.getGangMatchPolicy() == extension.GangMatchPolicyOnceSatisfied && gang.isGangOnceResourceSatisfied() { return nil }
Why is this needed:
The gang policy immediately becomes invalid after the pod is destroyed and rebuilt. The newly created pod becomes running even when the total number of pods does not meet the MinRequiredNumber. However, some users think this does not meet their expectations. They believe that the semantics of the gang policy are still followed in this scenario. , so it is recommended to add user intention judgment logic
Is there a suggested solution, if so, please add it:
Annotation can be added to the podGroup, for example: gang.scheduling.koordinator.sh/close-history-evaluate. At this time, the rebuilt pod is destroyed. When the total number of pods does not meet the MinRequiredNumber, it is still pending, maintaining gang semantics.
ps:If the community thinks this capability suggestion is reasonable, I am willing to submit the corresponding PR.
The text was updated successfully, but these errors were encountered: