[
https://issues.apache.org/jira/browse/PHOENIX-1677?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
James Taylor updated PHOENIX-1677:
----------------------------------
Attachment: PHOENIX-1677_v2.patch
Simpler version that hopefully solves the issue. Please let me know,
[~mujtabachohan], and if you can put together a unit test for this, that would
be much appreciated.
[~lhofhansl] - would you mind reviewing? We'll end up just defaulting to what
HBase does or what our ProtectedHConnection does at SFDC.
> Immutable index deadlocks when number of guideposts are one half of thread
> pool size
> ------------------------------------------------------------------------------------
>
> Key: PHOENIX-1677
> URL: https://issues.apache.org/jira/browse/PHOENIX-1677
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 5.0.0, 4.3
> Reporter: Mujtaba Chohan
> Assignee: James Taylor
> Attachments: PHOENIX-1677_v2.patch
>
>
> If total number of parallel iterators + mutation count exceeds Phoenix thread
> pool size then immutable index remains at 0 rows with no activity on client
> and server.
> Ex. Only with guide post count of 64 or lower does immutable index gets build
> with default 128 thread pool. By changing guide post width to get with >64
> guideposts, index fails to build. Also if thread pool size is lowered from
> 128 then index fails to build for 64 guide posts.
> Let me know if you need a utest as I think it would be easy to repro in it
> too.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
