Wei-Chiu Chuang created HDDS-17624:
--------------------------------------
Summary: Complete upgrade and downgrade documentation for Recon,
S3 Gateway, and HttpFS Gateway
Key: HDDS-17624
URL: https://issues.apache.org/jira/browse/HDDS-17624
Project: Apache Ozone
Issue Type: Bug
Reporter: Wei-Chiu Chuang
The Upgrade and Downgrade guide instructs operators to stop and replace all
components, but its startup procedure covers only SCM, datanodes, and OM. It
omits deployed Recon, S3 Gateway, and HttpFS Gateway services.
Guide:
https://ozone.apache.org/docs/administrator-guide/operations/upgrade-and-downgrade/
The guide also presents pre-finalization downgrade as a cluster-wide guarantee
without explaining Recon's independent metadata upgrade lifecycle. In Ozone
2.2.1, ReconServer calls ReconLayoutVersionManager.finalizeLayoutFeatures()
automatically on startup, independently of OM/SCM finalization.
This behavior is intentional: the description of HDDS-15374 / PR #10443
explicitly says Recon finalizes automatically on startup and preserves that
behavior. This issue requests accurate operational documentation, not a change
to that design.
https://github.com/apache/ozone/pull/10443
https://github.com/apache/ozone/blob/ozone-2.2.1/hadoop-ozone/recon/src/main/java/org/apache/hadoop/ozone/recon/ReconServer.java#L169-L179
Requested changes:
* Include Recon, S3 Gateway, and HttpFS Gateway in stop, binary replacement,
restart, and downgrade instructions when deployed.
* Document normal startup commands: ozone --daemon start recon, ozone --daemon
start s3g, and ozone --daemon start httpfs. Clarify that OM's
--upgrade/--downgrade flags do not apply to these services.
* Explain Recon's automatic startup migrations and independence from OM/SCM
finalization.
* Define supported Recon downgrade behavior, including release-specific
compatibility restrictions and any required backup, restore, or metadata
rebuild procedure. Do not assume all Recon state is disposable.
* Clarify the scope of the pre-finalization downgrade guarantee.
* Add checks for Recon health/synchronization and gateway functionality.
* Cross-reference
https://ozone.apache.org/docs/administrator-guide/operations/start-and-stop/
Acceptance criteria: An operator can follow the guide for a deployment
containing all six services, understand when Recon metadata changes, and
determine the supported recovery steps before starting an older Recon binary.
Recon downgrade compatibility should be investigated separately; automatic
migration alone does not establish a code defect.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
