[
https://issues.apache.org/jira/browse/HBASE-16296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15402468#comment-15402468
]
Andrew Purtell commented on HBASE-16296:
----------------------------------------
The client view of the scanning result is the same, but an early out avoids a
~200ms stall caused by extra unnecessary work and IO in the reverse scanning
case. This is a performance bug impacting us where I work so would like to
commit to move forward, unless objection.
We can fix FilterList separately with HBASE-16310
Let me make Ted's proposal into a patch, check that it fixes the perf problem,
check that it doesn't fail any tests, and then commit to master, and branch-1.
Separately, for 0.98, I will look at committing Lars' patch v2 using the same
criteria. Ted's patch relies entirely on code that doesn't exist in 0.98. Will
assign the issue to Ted when resolving.
> Reverse scan performance degrades when scanner cache size matches page filter
> size
> ----------------------------------------------------------------------------------
>
> Key: HBASE-16296
> URL: https://issues.apache.org/jira/browse/HBASE-16296
> Project: HBase
> Issue Type: Bug
> Components: Filters
> Reporter: James Taylor
> Fix For: 2.0.0, 1.3.0, 1.4.0, 1.1.6, 0.98.21, 1.2.3
>
> Attachments: 16296-MAYBE-v2.txt, 16296-MAYBE.txt, PAgefilter.patch,
> generatedata-snippet.java, repro-snippet.java
>
>
> When a reverse scan is done, the server seems to not know it's done when the
> scanner cache size matches the number of rows in a PageFilter. See
> PHOENIX-3121 for how this manifests itself. We have a standalone, pure HBase
> API reproducer too that I'll attach (courtesy of [~churromorales] and
> [~mujtabachohan]).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
