[
https://issues.apache.org/jira/browse/SPARK-36236?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
L. C. Hsieh reassigned SPARK-36236:
-----------------------------------
Assignee: Venki Korukanti
> RocksDB state store: Add additional metrics for better observability into
> state store operations
> ------------------------------------------------------------------------------------------------
>
> Key: SPARK-36236
> URL: https://issues.apache.org/jira/browse/SPARK-36236
> Project: Spark
> Issue Type: Sub-task
> Components: Structured Streaming
> Affects Versions: 3.1.2
> Reporter: Venki Korukanti
> Assignee: Venki Korukanti
> Priority: Major
>
> Proposing adding following new metrics to {{customMetrics}} under the
> {{stateOperators}} in {{StreamingQueryProgress}} event These metrics help
> have better visibility into the RocksDB based state store in streaming jobs.
> * {{rocksdbGetCount}} number of get calls to the DB (doesn’t include Gets
> from WriteBatch - in memory batch used for staging writes)
> * {{rocksdbPutCount}} number of put calls to the DB (doesn’t include Puts to
> WriteBatch - in memory batch used for staging writes)
> * {{rocksdbTotalBytesReadByGet/rocksdbTotalBytesWrittenByPut}}: Number of
> uncompressed bytes read/written by get/put operations
> * {{rocksdbReadBlockCacheHitCount/rocksdbReadBlockCacheMissCount}} indicates
> how much of the block cache in RocksDB is useful or not and avoiding local
> disk reads
> *
> {{rocksdbTotalBytesReadByCompaction/rocksdbTotalBytesWrittenByCompaction}}:
> How many bytes the compaction process read from disk and written to disk.
> * {{rocksdbTotalCompactionTime}}: Time (in ns) took for compactions (both
> background and the optional compaction initiated during the commit)
> * {{rocksdbWriterStallDuration}} Time (in ns) the writer has stalled due to
> a background compaction or flushing of the immutable memtables to disk.
> * {{rocksdbTotalBytesReadThroughIterator}} Some of the stateful operations
> (such as timeout processing in FlatMapGroupsWithState and watermarking)
> requires reading entire data in DB through iterator. This metric tells the
> total size of uncompressed data read using the iterator.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
