Recover an Instance
During update or upgrade operations, because of various issues on the provider or service side, instances are in unsuccessful states.
The following table explains the type of issue and respective states.
| Root Cause | Recoverable (Update or Upgrade) | Nonrecoverable (Create or Delete) |
|---|---|---|
|
Provider issue (Because of unsuccessful orchestration of building blocks in provider tenancy caused by problems on the provider side. For example, missing policies, hitting tenancy limit for building blocks.) |
RECOVERABLE_PROVIDER_ISSUE | NON_RECOVERABLE_PROVIDER_ISSUE |
| Service issue(Because of issues in OCI. This involves issues (temporary outage or defect in ML Applications Service or in some of its dependencies such as services providing building block resources.) | RECOVERABLE_SERVICE_ISSUE | NON_RECOVERABLE_SERVICE_ISSUE |
In the case of issues with providers, the issues need to be identified and fixed. After
fixing the issues, providers can call the recover action when the ML Application Instance view
is in RECOVERABLE_SERVICE_ISSUE or RECOVERABLE_PROVIDER_ISSUE state. This runs the last failed
update or upgrade operation.
| Endpoint | Related Resource | Required Permission | Notes |
|---|---|---|---|
/mlApplicationInstanceViews/{mlApplicationInstanceViewId}/actions/recover
|
ML Application Instance View | DATA_SCIENCE_APPLICATION_INSTANCE_VIEW_UPDATE | To be used by providers |
The recover action isn't supported for unsuccessful create or delete operations.