[
https://issues.apache.org/jira/browse/HBASE-10169?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Gary Helmling updated HBASE-10169:
----------------------------------
Attachment: HBASE-10169-alternate.patch
After the review I posted on the reviewboard request last week, I went through
the patch and made a number of changes in the direction I had suggested. My
primary comments on the v5 patch were:
{quote}
Thanks for all the work on this updated patch. The new client APIs seem
reasonable, though I think they could use a minor simplification.
My biggest feedback is that MultiRegionServerCallable and
RegionServerCoprocessorRpcInvoker seem to duplicate a lot of what is provided
by the existing MultiServerCallable and AsyncProcess. When I originally looked
at making this same change in a previous version of HBase, I looked at
extending the Action and MultiAction classes to support coprocessor endpoint
invocations so that we could leverage the existing RPC mechanisms. I still
think that would be a better approach here, rather that creating a parallel set
of RPC classes specifically for coprocessor endpoints. But if you disagree,
please just explain what problems you see with this approach.
{quote}
Attached is a modified patch (HBASE-10169-alternate.patch), which modifies the
code to use AsyncProcess as suggested. The main differences from the v5 patch
are:
* In HTable batchCoprocessorService methods, replaces ServiceDescriptor +
String method name with MethodDescriptor
* Removes MultiRegionServerCallable and RegionServerCoprocessorRpcInvoker,
using AsyncProcess instead
* Removes HRegionServer.execMultiService() method
* Removed the additional thread pool in HRegionServer. If this is useful to
have, we could add it back in in a separate issue (and consider doing so for
other batch operations at the same time).
* Adds support for CoprocessorServiceCall to Action/MultiAction, and
CoprocessorServiceResult to MultiResponse/ResultOrException
[[email protected]], please look through the alternate patch and let me
know if this approach might work for you, or if you still see some needed
functionality missing.
Note: TestBatchCoprocessorEndpoint#testAggregationWithNullResponse seems to be
failing occasionally. I'm still trying to track that down.
> Batch coprocessor
> -----------------
>
> Key: HBASE-10169
> URL: https://issues.apache.org/jira/browse/HBASE-10169
> Project: HBase
> Issue Type: Sub-task
> Components: Coprocessors
> Affects Versions: 0.99.0
> Reporter: Jingcheng Du
> Assignee: Jingcheng Du
> Attachments: Batch Coprocessor Design Document.docx,
> HBASE-10169-V2.patch, HBASE-10169-V3.patch, HBASE-10169-V3.patch,
> HBASE-10169-V4.patch, HBASE-10169-V5.patch, HBASE-10169-alternate.patch,
> HBASE-10169.patch
>
>
> This is designed to improve the coprocessor invocation in the client side.
> Currently the coprocessor invocation is to send a call to each region. If
> there’s one region server, and 100 regions are located in this server, each
> coprocessor invocation will send 100 calls, each call uses a single thread in
> the client side. The threads will run out soon when the coprocessor
> invocations are heavy.
> In this design, all the calls to the same region server will be grouped into
> one in a single coprocessor invocation. This call will be spread into each
> region in the server side.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
