[
https://issues.apache.org/jira/browse/HADOOP-18845?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18058222#comment-18058222
]
Steve Loughran commented on HADOOP-18845:
-----------------------------------------
ooh, looks similar enough doesn't it.
org.apache.hadoop.fs.s3a.impl.ErrorTranslation should catch the shaded and
unshaded versions of those (string match) and treat as a connectivity problem
and retry. Are you not seeing that?
it could be caused by
* far end not actually responding to the request for some reason
* client doing a put/post/get/head and then when it waits for a response, not
getting one. That is: it only notices the connection is stale by a failure to
get a response.
so yes, look at the ttls.
but we should be retrying on this, so check the stack trace and see why it's not
> Add ability to configure ConnectionTTL of http connections while creating S3
> Client.
> ------------------------------------------------------------------------------------
>
> Key: HADOOP-18845
> URL: https://issues.apache.org/jira/browse/HADOOP-18845
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
> Affects Versions: 3.3.6
> Reporter: Mukund Thakur
> Assignee: Mukund Thakur
> Priority: Major
> Labels: pull-request-available
> Fix For: 3.3.9, 3.5.0
>
>
> The option fs.s3a.connection.ttl sets the maximum time an idle connection may
> be retained in the http connection pool.
> A lower value: fewer connections kept open, networking problems related to
> long-lived connections less likely
> A higher value: less time spent negotiating TLS connections when new
> connections are needed
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
