[jira] [Commented] (HTTPCLIENT-2039) Do not close ConnectionManager in case of Errors

["Sean Guo (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HTTPCLIENT-2039?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17172051#comment-17172051 ] 

Sean Guo commented on HTTPCLIENT-2039:
--------------------------------------

We met the same issue in our product. It's kinda normal case for a OOM to occur in our processing logic and we will handle that according and recover. With this change it seems we have to recreate the client factory whenever this happens as we shared the same connection manager for all clients. 

Is it possible to control this breaking change via a configuration? Thanks.

> Do not close ConnectionManager in case of Errors
> ------------------------------------------------
>
>                 Key: HTTPCLIENT-2039
>                 URL: https://issues.apache.org/jira/browse/HTTPCLIENT-2039
>             Project: HttpComponents HttpClient
>          Issue Type: Improvement
>          Components: HttpClient (classic)
>    Affects Versions: 4.5.10
>            Reporter: Thomas Sartissohn
>            Priority: Minor
>
> The MainClientExec closes the connectionManager in the execute method whenever a java error is received. In my case an OutOfMemoryError was thrown. This ended in a client instance no longer usable cause the pool was already shut down. And therefore the real error (OutOfMemoryError) was covered by many "Connection pool shut down" exceptions.
> I agree that the OutOfMemoryError  is a severe issue which should be covered somehow. But closing the pool is unexpected here.
> Is there a chance to remove the closing part and keep the pool running in error case?



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org