[
https://issues.apache.org/jira/browse/HBASE-10357?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Devaraj Das updated HBASE-10357:
--------------------------------
Attachment: 10357-4.txt
Thanks for the reviews earlier [~stack], [~enis] and [~nkeywal].
Attaching a patch that should take care of most feedback. In this patch, I bit
the bullet and used the replica code path for all scans (with or without
replicas, STRONG or TIMELINE). That way it'd be tested well. Also, had to
refactor a little bit of ClusterConnection's replica methods and make it part
of public interface since I needed to access replica methods in in a unit test.
Left the default value for replica-timeout as 1 second. I think this is decent
for scans (there are two RPCs involved in the beginning for example). The
switching of replicas is per RPC call from the ScannerCallable's call method
([~enis], by batch are you referring to this per RPC request/response or the
whole scan?).
I'll try to come up with a unit test for the switching of replicas between a
scan but it'd require some hackery with coprocessors. Let me see if i can do it
cleanly. But otherwise, this patch should be close...
> Failover RPC's for scans
> ------------------------
>
> Key: HBASE-10357
> URL: https://issues.apache.org/jira/browse/HBASE-10357
> Project: HBase
> Issue Type: Sub-task
> Components: Client
> Reporter: Enis Soztutar
> Fix For: 0.99.0
>
> Attachments: 10357-1.txt, 10357-2.txt, 10357-3.2.txt, 10357-3.txt,
> 10357-4.txt
>
>
> This is extension of HBASE-10355 to add failover support for scans.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
