[jira] [Commented] (HBASE-21684) Throw DNRIOE when connection or rpc client is closed

["Duo Zhang (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-21684?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16735654#comment-16735654 ] 

Duo Zhang commented on HBASE-21684:
-----------------------------------

When implementing HBASE-21671, I found that when shutting down a RS, my new RegionReplicaReplicationEndpoint which is implemented with async client will retry a lot of times and generate bunch of error logs. Then I notice that, the problem is that, the async connection of the region server has been closed, but the replication endpoint is still trying to replicate entries, and since for region replica replication, we are replicating to the same cluster, so I just reuse the async connection of the region server, which has already been closed. The problem here is that, the StoppedRpcClientException is not a DNRIOE, so we will keep retrying...

So I suggest that we change StoppedRpcClientException to extends DNRIOE, and also for other places in hbase-client module, if we notice that the connection is closed, just throw a DNRIOE out.

> Throw DNRIOE when connection or rpc client is closed
> ----------------------------------------------------
>
>                 Key: HBASE-21684
>                 URL: https://issues.apache.org/jira/browse/HBASE-21684
>             Project: HBase
>          Issue Type: Improvement
>          Components: asyncclient, Client
>            Reporter: Duo Zhang
>            Priority: Major
>




--
This message was sent by Atlassian JIRA
(v7.6.3#76005)