✨ add ttlSeconds field to RayClusterSpec #4036
Conversation
danielgafni
requested review from
kevin85421,
andrewsykim,
rueian and
MortalHappiness
as code owners
| We use [elastic/crd-ref-docs](https://github.com/elastic/crd-ref-docs) to generate API reference for CRDs of KubeRay. The configuration file of `crd-ref-docs` is located at `hack/config.yaml`. Please refer to the documentation for more details. | ||
|
|
||
| Generate API refernece: | ||
| Generate API reference: |
There was a problem hiding this comment.
Thanks!
Just a friendly reminder that the auto-generated docs and Helm chart might need to be updated as well. You can usually regenerate them by running the commands in https://github.com/ray-project/kuberay/blob/master/ray-operator/DEVELOPMENT.md#consistency-check.
| // This field is set when TTLSeconds is specified in the RayClusterSpec. | ||
| // +nullable | ||
| // +optional | ||
| TTLExpirationTime *metav1.Time `json:"ttlExpirationTime,omitempty"` |
There was a problem hiding this comment.
Not sure we need this, can we just use creationTimestamp + ttlSeconds? Or is the TTL timer start once the RayCluster is ready?
| } | ||
|
|
||
| ttlSeconds := *instance.Spec.TTLSeconds | ||
| creationTime := instance.CreationTimestamp.Time |
There was a problem hiding this comment.
Should expiration time be from RayCluster creation or RayCluster ready state?
|
I have some questions about the behavior: https://github.com/ray-project/kuberay/issues/4033#issuecomment-3247488390. We should reach a conclusion before moving forward with this PR. |
Why are these changes needed?
Related issue number
Closes #4033
Checks