[
https://issues.apache.org/jira/browse/HBASE-24506?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17127496#comment-17127496
]
Bharath Vissapragada commented on HBASE-24506:
----------------------------------------------
> we will do other things which will acquire the lock of a RpcConnection again.
[~zhangduo] Can you please add a snippet of code that acquires the lock in
reverse? I skimmed through the callbacks in the code and I don't see a place
the lock acquisition is reverse (hrc first and connection second). I should've
figured it from the attached jstack, but it is incomplete. It only shows two
threads that are blocked in sendRequest().
> async client deadlock
> ---------------------
>
> Key: HBASE-24506
> URL: https://issues.apache.org/jira/browse/HBASE-24506
> Project: HBase
> Issue Type: Bug
> Components: asyncclient
> Affects Versions: 3.0.0-alpha-1, 2.3.0, 2.2.4, 2.4.0, 2.2.5
> Reporter: Whitney Jackson
> Assignee: Duo Zhang
> Priority: Blocker
> Fix For: 3.0.0-alpha-1, 2.3.0, 2.2.6
>
>
> I ran into an issue where one instance in a cluster of application servers
> backed by HBase stopped processing requests. Looking at a thread dump, it
> seems HBase client threads are deadlocked:
> https://pastebin.com/raw/B3FJL1Ag
> The deadlock seemed to happen at the same time that a region server was
> abruptly stopped (the physical server was restarted unexpectedly).
> I'm using the hbase async API. The hbase client version is 2.2.4. The server
> is running 2.2.4 as well.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
