StatefulSet Support for Updating Volume Claim Template #1
Conversation
huww98
commented
May 18, 2024
huww98
commented
- One-line PR description:
- Issue link:
- Other comments:
| e.g.: | ||
| - If `spec.updateStrategy.type` is `RollingUpdate`, | ||
| update the PVCs in the order from the largest ordinal to the smallest. | ||
| Only proceed to the next ordinal when all the PVCs of the previous ordinal |
There was a problem hiding this comment.
Does just being compatible mean we can move on to the next ordinal? Shouldn't the resizing be finished?
There was a problem hiding this comment.
By saying compatible, we are already waiting for the resize to finish. Because in "What PVC is compatible", status.capacity.storage should be updated before a PVC becomes compatible.
| - totalCapacity: the sum of `status.capacity` of all the PVCs. | ||
|
|
||
| Some fields in the `status` are also updated to reflect the staus of the PVCs: | ||
| - readyReplicas: in addition to pods, also consider the PVCs status. A PVC is not ready if: |
There was a problem hiding this comment.
If strategy is ondelete, and old pvc works, the corresponding pod is ready or not?
There was a problem hiding this comment.
if volumeClaimUpdateStrategy is OnDelete, the PVC is always considered ready. This is required to not break the current behavior.
| (`spec.resources.requests.storage`, `spec.volumeAttributesClassName`, `metadata.labels`, and `metadata.annotations` currently), | ||
| update the PVC in-place to the extent possible. | ||
| Do not perform the update that will be rejected by API server, such as | ||
| decreasing the storage size below its current status. |
There was a problem hiding this comment.
Change API server to allow any updates to
volumeClaimTemplateof a StatefulSet.
I'm confused. Will decreasing the storage size be supported in this proposal?
There was a problem hiding this comment.
Decreasing in template is supported, but only for the purpose of recovering from failed expansion, or as template for new PVCs. Shrink the PV/PVC is not supported.
We accept any update to the template, but decreasing the size will not affect the existing healthy PVC.
| - `volumeClaimUpdateStrategy` is `InPlace` and the PVC is updating; | ||
| - availableReplicas: total number of replicas of which both Pod and PVCs are ready for at least `minReadySeconds` | ||
| - currentRevision, updateRevision, currentReplicas, updatedReplicas | ||
| are updated to reflect the status of PVCs. |
There was a problem hiding this comment.
When expanding the size of pvcs and then kubectl rollout undo, actually pvcs will not be rollbacked, they will stay the expanded size. Right?
There was a problem hiding this comment.
It depends. If you enabled feature gate RecoverVolumeExpansionFailure, and the PVC pvc.Status.AllocatedResources is lower, then PVC is rolled back, and expansion is cancelled.
This is what I mean by writing "update the PVC in-place to the extent possible". Maybe I should make this more explicit.
|
|
||
| #### Story 5: Asymmetric Replicas | ||
|
|
||
| The replicas of our StatefulSet are not identical, so we still want to update |
There was a problem hiding this comment.
| The replicas of our StatefulSet are not identical, so we still want to update | |
| The templates for different replicas of a StatefulSet are not identical, so we still want to update |
There was a problem hiding this comment.
I would like the word "template" to specifically refer to the one written in StatefulSet. So I don't like "templates for different replicas". Maybe "The storage requirement of different replicas"?
|
|
||
| We're running a CI/CD system and the end-to-end automation is desired. | ||
| To expand the volumes managed by a StatefulSet, | ||
| we can just use the same pipeline that we are already using to updating the Pod. |
There was a problem hiding this comment.
| we can just use the same pipeline that we are already using to updating the Pod. | |
| we can just use the same pipeline that we are already using to update the Pod. |
| 2. Introduce a new field in StatefulSet `spec.updateStrategy.rollingUpdate`: `volumeClaimSyncStrategy` | ||
| to specify how to update PVCs and Pods. Possible values are: | ||
| - `Async`: the default value, preseve the current behavior. | ||
| - `LockStep`: update PVCs first, then update Pods. See below for details. |
There was a problem hiding this comment.
I don't understand why it's called LockStep. I came up with a few names for you to consider. Stepwise, Serial, Ordered, Sync.
There was a problem hiding this comment.
See here for the context of word lock-step.
