[
https://issues.apache.org/jira/browse/CASSANDRA-14149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17941078#comment-17941078
]
Yuqi Yan commented on CASSANDRA-14149:
--------------------------------------
> Currently, one range tombstone will count for 2 tombstone cells but as they
> may shadow thousands if not millions of actual cells, some latency issues in
> the local read path may be hard to diagnose.
Was in the exact same situation and it took lots of efforts understanding why
the local read was so slow..
Adding table level metrics that collect such shadowed rows information when
merging the rows from different sources
([PR|https://github.com/apache/cassandra/pull/4056])
> Count all types of tombstones in the read path and the rows/cells they shadow
> -----------------------------------------------------------------------------
>
> Key: CASSANDRA-14149
> URL: https://issues.apache.org/jira/browse/CASSANDRA-14149
> Project: Apache Cassandra
> Issue Type: Improvement
> Components: Legacy/Local Write-Read Paths
> Reporter: Alexander Dejanovski
> Assignee: Yuqi Yan
> Priority: Normal
> Fix For: 5.x
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> This ticket is a follow up to CASSANDRA-8527 which aims at reporting row
> level tombstones in the read path.
> The aim here is to fully categorize the cells/rows that can be shadowed by
> range tombstones. Currently, one range tombstone will count for 2 tombstone
> cells but as they may shadow thousands if not millions of actual cells, some
> latency issues in the local read path may be hard to diagnose.
> Based on discussions on the ML, one outcome could be to distinctively count
> cell/row/range tombstones and the number of cells/rows shadowed by each type.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
