[
https://issues.apache.org/jira/browse/HBASE-24024?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17110382#comment-17110382
]
Viraj Jasani commented on HBASE-24024:
--------------------------------------
[~mokai87] scan is different from multi puts/deletes. In multi(), user provides
rows(cells) to update in batch request but scan() scans through all records of
the table. You can provide limit and various filters but I am not sure if there
should be any restrictions on caching while performing scan.
In fact, this is good point, we have also observed scans causing very high
hotspots to specific region and high GC on that RegionServer. However, limiting
anything on cache level is something we should consider while not impacting
scan performance / client queries.
[~anoop.hbase] [~reidchan] thought?
> Optionally reject multi() requests with very high no of rows
> ------------------------------------------------------------
>
> Key: HBASE-24024
> URL: https://issues.apache.org/jira/browse/HBASE-24024
> Project: HBase
> Issue Type: Improvement
> Reporter: Viraj Jasani
> Assignee: Viraj Jasani
> Priority: Major
> Fix For: 3.0.0-alpha-1, 2.3.0, 1.7.0
>
>
> As part of multi() request, RegionServer might receive requests with higher
> rows than row size threshold defined by hbase.rpc.rows.warning.threshold, and
> we provide warning log. To improve CPU consumption, we should introduce a
> config that would let RegionServer reject such requests other than providing
> warning log. This is an attempt to not let any client overburden CPU
> computation available to RegionServers.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
