[
https://issues.apache.org/jira/browse/HBASE-22135?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16806057#comment-16806057
]
Duo Zhang commented on HBASE-22135:
-----------------------------------
As now the AsyncMasterRequestRpcRetryingCaller inherits AsyncRpcRetryingCaller,
I think we can reuse the retry logic in it, which means we do not need to call
isMasterRunning any more, just cache the master address and use it, once we get
an error, we reset the master address cache and fetch it from zk again.
Let me try.
> AsyncAdmin will not refresh master address
> ------------------------------------------
>
> Key: HBASE-22135
> URL: https://issues.apache.org/jira/browse/HBASE-22135
> Project: HBase
> Issue Type: Sub-task
> Components: Admin, asyncclient, Client
> Reporter: Duo Zhang
> Priority: Major
>
> This is a big problem... If we stop the active master and promote the backup
> master, all methods in AsyncAdmin, which need to connect the master will fail
> as they try to connect to the old active master forever...
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
