[
https://issues.apache.org/jira/browse/PHOENIX-3783?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15974064#comment-15974064
]
Lars Hofhansl edited comment on PHOENIX-3783 at 4/19/17 4:55 AM:
-----------------------------------------------------------------
Batches that update global, mutable indexes, right? Of course that's perfectly
reasonable. :)
So something 2MB or 1000 rows, whatever is less?
Don't want to make the batches too small in terms of bytes. What's expensive
are RPC calls with very little payload (it's OK to eat an RTT for 2MB, but it's
very inefficient to waste an RTT on a few KBs).
was (Author: lhofhansl):
Batches that update global, mutable indexes, right? Of course that's perfectly
reasonable. :)
So something 2MB or 1000 rows, whatever is less?
> Lower max mutation bytes to create smaller MutationState chunks
> ---------------------------------------------------------------
>
> Key: PHOENIX-3783
> URL: https://issues.apache.org/jira/browse/PHOENIX-3783
> Project: Phoenix
> Issue Type: Improvement
> Affects Versions: 4.10.0
> Reporter: Geoffrey Jacoby
> Assignee: Geoffrey Jacoby
> Attachments: PHOENIX-3783.patch
>
>
> In PHOENIX-541 we added a feature to break large MutationStates into smaller
> chunks so that they could be submitted to HBase without causing performance
> issues. The default max byte size we originally chose has proven much too
> high and should be lowered.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
