Siyao Meng created HDDS-16695:
---------------------------------
Summary: Closing any snapshot DB unregisters the process wide
RocksDB metrics source
Key: HDDS-16695
URL: https://issues.apache.org/jira/browse/HDDS-16695
Project: Apache Ozone
Issue Type: Bug
Reporter: Siyao Meng
Attachments: HDDS-16695-CR-R12-shared-rocksdb-metrics-source.patch
h3. Mechanism
{{RDBStore.close()}} calls the static {{RDBMetrics.unRegister()}}
({{RDBStore.java:242-268}}, {{RDBMetrics.java:126-130}}), which nulls the
process wide singleton and unregisters the shared "Rocks DB Metrics" source.
The active OM DB and every snapshot DB are {{RDBStore}}s in one JVM sharing
that one source, and nothing reference counts it.
h3. Impact
Closing any snapshot DB tears down the RocksDB metrics source shared with the
active OM DB, so the JMX counters an operator trends reset to zero. Snapshot
closes happen routinely, from {{SnapshotCache}} eviction
({{SnapshotCache.java:407}}) and purge time invalidate ({{:161}}). Metrics
only, with no data or availability impact.
h3. Reproduction
{{repro/test_bugCR-R12_shared_rocksdb_metrics.sh}}. Evidence in
{{repro/logs/ev-cr2.log}}.
h3. Patch
{{HDDS-XXXXX-CR-R12-shared-rocksdb-metrics-source.patch}} is attached. It
applies to 20e6a1c0f6039248373efd3eb1d7f2afcf7f1535 and carries the fix plus a
regression test that was verified to fail without the fix and pass with it.
{code:none}
mvn test -pl hadoop-hdds/framework -Dtest=TestRDBMetrics
{code}
Found by code review of the OM snapshot cache under HDDS-15926, on commit
20e6a1c0f6039248373efd3eb1d7f2afcf7f1535. Generated with Specula (Claude Opus
5).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
