Add Leader Election Resource Lock Options #6426
Conversation
There was a problem hiding this comment.
@IzzyMusa Thanks for this contribution! I think the change makes sense and looks good to me. Only change needed for my approval is fixing the changelog
changelog/fragments/00-template.yaml
Outdated
There was a problem hiding this comment.
This is a template file that should not be changed. Instead, would you mind copying the contents of this file into a new file and modifying it to reflect the changes you are making in this PR?
There was a problem hiding this comment.
@everettraven Gotcha! Thanks!
There was a problem hiding this comment.
@everettraven Do you know which version of OpenShift Operator-SDK will support in v1.29.0? If 4.13 I might as well change the resource lock to be lease here
There was a problem hiding this comment.
@IzzyMusa I think the version of operator-sdk that got pulled down for 4.13 was v1.28.? (not sure if a patch version went in or not)
There was a problem hiding this comment.
@everettraven I updated the change log. Yup, based on this here. I can't find the exact place where I saw the migration plan from configmap to leases but I also added a flag for the type of resources lock. I can default it to configmapLeases if you think folks with older clusters will still depend on the configmap as a resource lock obj.
There was a problem hiding this comment.
Nit in the changelog, but other than that I think this looks good. Thanks @IzzyMusa !
changelog/fragments/add-leader-election-resource-lock-options.yaml
Outdated
Show resolved
Hide resolved
Signed-off-by: Catherine Chan-Tse <cchantse@redhat.com> Signed-off-by: Izzy Musa <izzy.musa@ibm.com>
…erator Signed-off-by: Izzy Musa <izzymusa@Izzys-MacBook-Pro.local> Signed-off-by: Izzy Musa <izzy.musa@ibm.com>
Signed-off-by: Izzy Musa <izzymusa@Izzys-MacBook-Pro.local> Signed-off-by: Izzy Musa <izzy.musa@ibm.com>
Signed-off-by: Izzy Musa <izzy.musa@ibm.com>
Signed-off-by: Izzy Musa <izzy.musa@ibm.com>
This reverts commit 443be19. Signed-off-by: Izzy Musa <izzy.musa@ibm.com>
Signed-off-by: Izzy Musa <izzy.musa@ibm.com>
Signed-off-by: Izzy Musa <izzy.musa@ibm.com>
This reverts commit 29f8b9e. Signed-off-by: Izzy Musa <izzy.musa@ibm.com>
…yaml Co-authored-by: Bryce Palmer <everettraven@gmail.com> Signed-off-by: Izzy Musa <izzy.musa@ibm.com>
Signed-off-by: Izzy Musa <izzy.musa@ibm.com>
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
|
@everettraven There is a test that failed due to a timeout. Is it just a build error or do I need to change something? |
|
@IzzyMusa I'll re-run it but if it fails again I'll verify nothing is actually broken and override if everything checks out |
Description of the change:
Background:
Currently, the operator SDK for the Ansible operator lacks options to customize the leader election resource lock behavior. This limits the flexibility of the operator and prevents users from adjusting lease duration and renew deadline according to their specific needs.
Changes:
I made enhancements to the operator SDK to include additional flags for configuring leader election resource lock options. These changes provide users with the ability to specify the type of resource object used for locking during leader election and customize the lease duration and renew deadline.
Introduced
--leader-elect-lease-durationflag to allow users to define the duration that non-leader candidates will wait to force acquire leadership. The default duration is set to 15 seconds.Introduced
--leader-elect-renew-deadlineflag, enabling users to set the renew deadline, which determines the duration that the acting control plane will retry refreshing leadership before giving up. The default duration is set to 10 seconds.These changes provide more flexibility and control over leader election behavior, allowing operators to adapt to various deployment scenarios and specific requirements.
Reason for the Change:
The motivation behind these enhancements is to address frequent restarts observed in the operator instances. It was found that the leader election was often lost because the instance couldn't update the resource lock in time, resulting in the assumption that it crashed. By allowing users to adjust the lease duration and renew deadline, operators can mitigate the impact of resource lock updates, reducing unnecessary restarts and ensuring uninterrupted operation.
Expected Impact:
By incorporating these enhancements, operators built with the Ansible operator SDK will allow users to fine-tune the operators leader election resource locking.
Checklist
If the pull request includes user-facing changes, extra documentation is required:
changelog/fragments(seechangelog/fragments/00-template.yaml)website/content/en/docs