[
https://issues.apache.org/jira/browse/HDFS-14587?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16868853#comment-16868853
]
Wei-Chiu Chuang commented on HDFS-14587:
----------------------------------------
shouldn't it simply get a read timeout? i think we added a client side read
timeout not too long ago.
The only possibility I can imagine is the client gets into a full GC. But even
that, 9 hours seems like a stretch.
> Support fail fast when client wait ACK by pipeline over threshold
> -----------------------------------------------------------------
>
> Key: HDFS-14587
> URL: https://issues.apache.org/jira/browse/HDFS-14587
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: hdfs-client
> Reporter: He Xiaoqiao
> Assignee: He Xiaoqiao
> Priority: Major
>
> Recently, I meet corner case that client wait for data to be acknowledged by
> pipeline over 9 hours. After check branch trunk, I think this issue still
> exist. So I propose to add threshold about wait timeout then fail fast.
> {code:java}
> 2019-06-18 12:53:46,217 WARN [Thread-127] org.apache.hadoop.hdfs.DFSClient:
> Slow waitForAckedSeqno took 35560718ms (threshold=30000ms)
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
