[
https://issues.apache.org/jira/browse/HBASE-14489?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Lars Hofhansl updated HBASE-14489:
----------------------------------
Attachment: 14489-0.98.txt
Here's my patch from HBASE-10047 revived. Also added a test that ensures that
the hook is still called if a coprocessor implements it.
That brings a scan of a 10m rows table with 1 col and 1 version each (that has
been setup with a region observer that does not implement postScannerFilterRow)
from 7.9s back down to 3.0s (on a single machine).
We should also investigate why calling coprocessors is expensive, this should
lightweight enough to make a no-op call even for every row. I suspect is has to
do with the classloader checks we need to do.
[~apurtell], wanna have a look?
> postScannerFilterRow consumes a lot of CPU
> ------------------------------------------
>
> Key: HBASE-14489
> URL: https://issues.apache.org/jira/browse/HBASE-14489
> Project: HBase
> Issue Type: Bug
> Reporter: Lars Hofhansl
> Labels: performance
> Attachments: 14489-0.98.txt
>
>
> During an unrelated test I found that when scanning a tall table with CQ only
> and filtering most results at the server, 50%(!) of time is spend in
> postScannerFilterRow, even though the coprocessor does nothing in that hook.
> We need to find a way not to call this hook when not needed, or to question
> why we have this hook at all.
> I think [~ram_krish] added the hook (or maybe [~anoop.hbase]). I am also not
> sure whether Phoenix uses this hook ([~giacomotaylor]?)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
