[jira] [Commented] (CASSANDRA-11865) Improve compaction logging details

["Carl Yeksigian (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CASSANDRA-11865?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15644189#comment-15644189 ] 

Carl Yeksigian commented on CASSANDRA-11865:
--------------------------------------------

[~weideng]: This ticket is focused on adding new content to the {{compaction.log}}, not necessarily focused on the current debug logging.

Some of this data is already available, like the merge statistics, and most of the other data can be obtained using the {{MergeListener}}. For example, figuring out the number of merged CQL rows, as well as the merged tombstones.

The wide partition detection is done inside of {{BigTableWriter}}, so that will be more difficult to pull out, but we could probably get the same thing from going through {{CompactionAwareWriter}} instead of the {{BigTableWriter}}.

Metrics we already collect
* Partitions Processed: {{totalKeysWritten}} in {{CompactionTask}}
* Partition Merge Stats: {{mergedRowCounts}} in {{CompactionTask}}
* Partition Size: {{mean}} in {{StatsMetadata}} from the output SSTables

Metrics we need to collect
* Adding in {{MergeListener}} in {{CompactionIterator}}
** Rows Processed
** Rows Merged
** Row count min/max/avg per partition
** Cells Requiring Merge
** Number of Tombstones Encountered
* {{BigTableWriter}}/{{CompactionAwareWriter}}
** Wide row

> Improve compaction logging details
> ----------------------------------
>
>                 Key: CASSANDRA-11865
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-11865
>             Project: Cassandra
>          Issue Type: Sub-task
>          Components: Compaction
>            Reporter: T Jake Luciani
>            Assignee: Carl Yeksigian
>
> I'd like to see per compaction entry:
>   * Partitions processed
>   * Rows processed
>   * Partition merge stats
>   * If a wide row was detected
>   * The partition min/max/avg size
>   * The min/max/avg row count across partitions
> Anything else [~krummas]?



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)