Sammi Chen created HDDS-8325:
--------------------------------
Summary: Consolidate and refine RocksDB metrics
Key: HDDS-8325
URL: https://issues.apache.org/jira/browse/HDDS-8325
Project: Apache Ozone
Issue Type: Improvement
Reporter: Sammi Chen
Assignee: Sammi Chen
Attachments: Screenshot 2023-03-30 at 12-11-53 Screenshot.png,
Screenshot 2023-03-30 at 12-14-38 Screenshot.png, Screenshot 2023-03-30 at
12-21-48 Screenshot.png
RocksDB related metrics are important indicators of how RocksDB is running, and
also can facilitate RocksDB related issue investigation.
The majority of RocksDB metrics are exposed when property
"ozone.metastore.rocksdb.statistics" is set to 'ALL'.
Issues found are,
1. RocksDB metric of one RocksDB instance are exposed twice in jmx, one time
in prometheus. Here is DN example, dn-crl.db is exposed twice in JMX with
different metrics name. (b) is exposed in prometheus
a. ) "name" : "Hadoop:service=Ozone,name=RocksDbStore,dbName=dn-crl.db",
b. ) "name" : "Hadoop:service=HddsDatanode,name=Rocksdb_dn-crl.db",
2. blob related RocksDB metrics are exposed. Blob DB is not used in Ozone, we
can filter these parts of metrics
3. duplicate "estimated key count" info exposed in both Rocksdb_scm.db and
SCMMetadataStoreMetrics.
This task will
1. remove the duplicated exposure of RocksDB instance in jmx
2. filter blob db related metrics
3. remove SCMMetadataStoreMetrics as its function is covered by Rocksdb_scm.db
already.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
