[
https://issues.apache.org/jira/browse/FLINK-38867?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
James Kan updated FLINK-38867:
------------------------------
Description:
h4. Description
The FlinkBlueGreenDeployment controller does not properly honor the
[*savepointRedeployNonce*|https://nightlies.apache.org/flink/flink-kubernetes-operator-docs-main/docs/custom-resource/job-management/#redeploy-using-the-savepointredeploynonce]
parameter during subsequent deployments. When the nonce is incremented, the
controller ignores the user-specified *initialSavepointPath* and instead uses
the savepoint from the previous deployment (blue or green) being terminated.
h4. Expected Behavior
When *savepointRedeployNonce* is incremented in the FlinkBlueGreenDeployment
spec, the controller should respect the user-specified *initialSavepointPath*
and ensure the new active deployment (blue or green being transitioned to)
starts from the specified savepoint location.
h4. Current Behavior
- Initial deployments: *initialSavepointPath* is correctly applied ✓
- Subsequent deployments for FlinkBlueGreenDeployments: *initialSavepointPath*
is ignored when *savepointRedeployNonce* is updated ✗.
h4. Impact
This prevents manual savepoint selection functionality required for manual
recoveries and creates inconsistencies with the expected behaviour of
*savepointRedeployNonce* described in the Flink documentation.
was:
h4. Description
The FlinkBlueGreenDeployment controller does not properly honor the
[`savepointRedeployNonce`|https://nightlies.apache.org/flink/flink-kubernetes-operator-docs-main/docs/custom-resource/job-management/#redeploy-using-the-savepointredeploynonce]
parameter during subsequent deployments. When the nonce is incremented, the
controller ignores the user-specified `initialSavepointPath` and instead uses
the savepoint from the previous deployment (blue or green) being terminated.
h4. Expected Behavior
When `savepointRedeployNonce` is incremented in the FlinkBlueGreenDeployment
spec, the controller should respect the user-specified `initialSavepointPath`
and ensure the new active deployment (blue or green being transitioned to)
starts from the specified savepoint location.
h4. Current Behavior
- Initial deployments: `initialSavepointPath` is correctly applied ✓
- Subsequent deployments for FlinkBlueGreenDeployments: `initialSavepointPath`
is ignored when `savepointRedeployNonce` is updated ✗.
h4. Impact
This prevents manual savepoint selection functionality required for manual
recoveries and creates inconsistencies with the expected behaviour of
savepointRedeployNonce described in the Flink documentation.
> FlinkBlueGreenDeployment: savepointRedeployNonce does not respect
> user-provided initialSavepointPath
> ----------------------------------------------------------------------------------------------------
>
> Key: FLINK-38867
> URL: https://issues.apache.org/jira/browse/FLINK-38867
> Project: Flink
> Issue Type: Bug
> Components: Kubernetes Operator
> Reporter: James Kan
> Priority: Minor
>
> h4. Description
> The FlinkBlueGreenDeployment controller does not properly honor the
> [*savepointRedeployNonce*|https://nightlies.apache.org/flink/flink-kubernetes-operator-docs-main/docs/custom-resource/job-management/#redeploy-using-the-savepointredeploynonce]
> parameter during subsequent deployments. When the nonce is incremented, the
> controller ignores the user-specified *initialSavepointPath* and instead uses
> the savepoint from the previous deployment (blue or green) being terminated.
> h4. Expected Behavior
> When *savepointRedeployNonce* is incremented in the FlinkBlueGreenDeployment
> spec, the controller should respect the user-specified *initialSavepointPath*
> and ensure the new active deployment (blue or green being transitioned to)
> starts from the specified savepoint location.
> h4. Current Behavior
> - Initial deployments: *initialSavepointPath* is correctly applied ✓
> - Subsequent deployments for FlinkBlueGreenDeployments:
> *initialSavepointPath* is ignored when *savepointRedeployNonce* is updated ✗.
> h4. Impact
> This prevents manual savepoint selection functionality required for manual
> recoveries and creates inconsistencies with the expected behaviour of
> *savepointRedeployNonce* described in the Flink documentation.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
