lasdf1234 opened a new issue, #13964:
URL: https://github.com/apache/gravitino/issues/13964
### Version
main branch / 1.3.x
### Describe what's wrong
On multi-node (or multi-replica) startup, `IdpRESTFeature.configure` calls
`IdpUserGroupManager.initializeConfiguredServiceAdmins`. Each node checks
whether configured service admins exist, then inserts missing ones. When two
nodes both observe a missing admin and insert concurrently, one hits the unique
constraint on `(user_name, deleted_at)` and the duplicate-key error becomes
`AlreadyExistsException`. That exception currently fails Jetty startup of
`gravitino-web` (`UNAVAILABLE`), so the API can be unavailable until a later
restart succeeds.
Bootstrap should treat a concurrent create as success (idempotent): keep the
existence pre-check, and ignore `AlreadyExistsException` on insert. Do not
change the normal `addUser` API path.
Reported in production on enterprise `branch-1.3`
(datastrato/gravitino-enterprise#2723 / DAT-1704).
### Error message and/or stacktrace
```text
Caused by: org.apache.gravitino.exceptions.AlreadyExistsException: IdP user
admin already exists
at
org.apache.gravitino.idp.storage.relational.converters.IdpSQLExceptionConverter.toIdpException(...)
at
org.apache.gravitino.idp.storage.service.IdpUserMetaService.insertIdpUser(...)
at
org.apache.gravitino.idp.IdpUserGroupManager.initializeConfiguredServiceAdmins(...)
at
org.apache.gravitino.idp.web.rest.feature.IdpRESTFeature.configure(...)
Caused by: org.postgresql.util.PSQLException: ERROR: duplicate key value
violates unique constraint "idp_user_meta_user_name_deleted_at_key"
Detail: Key (user_name, deleted_at)=(admin, 0) already exists.
```
### How to reproduce
1. Deploy Gravitino with built-in IdP and configured service admins on a
shared PostgreSQL (or other) entity store.
2. Start two server nodes at the same time (empty IdP users, or after delete
of the admin row).
3. Observe one node may fail startup with `AlreadyExistsException` while the
other creates the admin successfully.
### Additional context
- Component: `plugins/idp-basic`
(`IdpUserGroupManager.initializeConfiguredServiceAdmins`)
- Fix approach: catch `AlreadyExistsException` only in the bootstrap insert
loop; leave user-facing create APIs unchanged.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
