[
https://issues.apache.org/jira/browse/HBASE-17719?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15892702#comment-15892702
]
James Moore commented on HBASE-17719:
-------------------------------------
Overall the issue is two fold
1) ScannerCallableWithReplicas creates a new RPCFactory for every call, which
will generate new interceptor factories and prevent the fast fail interceptor
from building up a map of failures over time. I've made the modifications for
clientScanner to pass along the rpcFactory to child scanners.
2) ScannerCallableWithReplicas/RetryingRPC does not extend RegionServerCallable
as ScannerCallable does, and won't pass the appropriate instanceof checks in
FastFailInterceptor. I've made the necessary modifications to allow the
Callables to pass the instanceof checks.
This patch is based off of branch-1.2 do I need to generate additional patch
files for 1.3/2.0?
--James
> PreEmptive Fast Fail does not apply to scanners
> -----------------------------------------------
>
> Key: HBASE-17719
> URL: https://issues.apache.org/jira/browse/HBASE-17719
> Project: HBase
> Issue Type: Bug
> Components: Client
> Affects Versions: 1.2.0
> Reporter: James Moore
> Assignee: James Moore
> Attachments: HBASE_17719.patch
>
>
> on CDH 5.9.0 testing revealed that scanners do not leverage Pre-emptive fast
> fail.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
