[
https://issues.apache.org/jira/browse/CASSANDRA-5366?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13607797#comment-13607797
]
Brooke Bryan commented on CASSANDRA-5366:
-----------------------------------------
The problem we have, is the amount of data within the node, we could have a few
large compactions run, before the upgrade process gets around to that specific
CF or SSTable, meaning we are re-processing some pretty large files (few
hundred GBs). An upgradesstables process is taking around 3 or 4 days to
complete at the moment, with the main problem being re-processing these large
sstables.
> UpgradeSSTables Optimisation
> ----------------------------
>
> Key: CASSANDRA-5366
> URL: https://issues.apache.org/jira/browse/CASSANDRA-5366
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Brooke Bryan
>
> Currently, if you run upgradesstables, cassandra will run through every
> single SSTable within the scope of the request. Where we have some large
> tables, an upgrade on a single sstable can take hours, even if its already
> sat on the same version.
> After upgrading to a new cassandra version, it would be ideal to be able to
> upgrade only sstables not sat in the latest version, as it seems like it just
> needs to do a massive amount of disk IO, with nothing being achieved at the
> end of it.
> Maybe its worth putting an option onto the nodetool command, or creating a
> new command for this type of upgrade
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
