[ https://issues.apache.org/jira/browse/CASSANDRA-9579?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15000464#comment-15000464 ]
Aleksey Yeschenko commented on CASSANDRA-9579: ---------------------------------------------- nodetool/JMX part LGTM. Changes to scheduling are somewhat unnecessarily overcomplicated. We don't really need to cancel the currently scheduled run if we got a new requests, it's no big deal. And if we assume it's no big deal, then the changes to {{SizeEstimatedRecorder}} can all be reverted. Pushed a simplified v2 [here|https://github.com/iamaleksey/cassandra/commits/9579-2.1], which is basically the v1, but with changes to {{SizeEstimatesRecorder}} reverted. > Add JMX / nodetool command to refresh system.size_estimates > ----------------------------------------------------------- > > Key: CASSANDRA-9579 > URL: https://issues.apache.org/jira/browse/CASSANDRA-9579 > Project: Cassandra > Issue Type: Improvement > Components: CQL > Reporter: Piotr Kołaczkowski > Assignee: Carl Yeksigian > Priority: Minor > Fix For: 2.1.x, 2.2.x > > > CASSANDRA-7688 added dumping size estimates at a fixed interval. However, in > some cases, e.g. after inserting huge amounts of data or truncating a table, > size estimates may become severely incorrect for the interval time. In this > case being able to manually trigger the recalculation of the estimates would > be very useful. It would be also useful for any automated testing requiring > fresh size estimates. -- This message was sent by Atlassian JIRA (v6.3.4#6332)