[
https://issues.apache.org/jira/browse/HBASE-13351?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14496790#comment-14496790
]
Josh Elser commented on HBASE-13351:
------------------------------------
Short answer: {{ConnectionUtils#createShortCircuitHConnection}} doesn't
actually do what it thinks its doing.
For callers which directly try to get the client service stub, things are fine.
However, in our specific case where we're constructing an HTable and calling a
Get on it, these methods are invoked on the ConnectionImplementation, not the
SCC stub. In other words, the HTable never even sees the SCC and always does a
normal RPC.
IMO, this is something to branch off into its own issue. The outstanding
question at hand here is admin threadpool sizing concerns for real instances,
of course this should likely also be done elsewhere :). I think the patch here
is ready to go, but we might need some thought about how requests actually get
routed since admin is already all meta requests + admin/server users (after
HBASE-13375) + hbase server-to-server RPCs (this ticket).
> Annotate internal MasterRpcServices methods with admin priority
> ---------------------------------------------------------------
>
> Key: HBASE-13351
> URL: https://issues.apache.org/jira/browse/HBASE-13351
> Project: HBase
> Issue Type: Improvement
> Components: master
> Reporter: Josh Elser
> Assignee: Josh Elser
> Fix For: 2.0.0, 1.1.0
>
> Attachments: HBASE-13351-v1.patch, HBASE-13351-v2.patch,
> HBASE-13351-v3.patch, HBASE-13351.patch
>
>
> HBASE-12071, among other things, introduced annotating RPC methods to give
> certain methods priority over others. Namely, this helps ensure that client
> requests cannot starve out internal RPC between master and regionserver.
> Similarly, we can do the same thing for Master RPC methods that are invoked
> by RS's.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
