Devesh Kumar Singh created HDDS-15308:
-----------------------------------------
Summary: Improve ICR/FCR-driven container state recovery using
SCM-authoritative state
Key: HDDS-15308
URL: https://issues.apache.org/jira/browse/HDDS-15308
Project: Apache Ozone
Issue Type: Task
Components: Ozone Recon
Affects Versions: 2.1.0
Reporter: Devesh Kumar Singh
Assignee: Devesh Kumar Singh
**Description**
Recon currently advances existing containers from `OPEN` to `CLOSING` when a
healthy non-`OPEN` DataNode replica report is received, but the DN-report path
does not fully converge several stale lifecycle states without waiting for the
periodic targeted SCM container sync.
Enhance Recon's ICR/FCR handling for existing containers so that:
- when Recon has `CLOSING`, Recon checks SCM's authoritative state and advances
to `QUASI_CLOSED`, `CLOSED`, `DELETING`, or `DELETED` when appropriate
- when Recon has `DELETED` but a DN reports a live replica and SCM still reports
`QUASI_CLOSED` or `CLOSED`, Recon rebuilds the local container entry from SCM
metadata instead of remaining permanently stale
- open-container pipeline accounting remains correct when moving
`OPEN -> CLOSING`
- focused unit/integration coverage is added for the above ICR/FCR recovery
cases
This follow-up should stay separate from HDDS-14758, whose scope is periodic
targeted SCM-to-Recon reconciliation.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
