[ 
https://issues.apache.org/jira/browse/QPID-7394?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Keith Wall updated QPID-7394:
-----------------------------
    Description: 
When the application calls {{Session#receive()}}, the client will automatically 
open the prefetch window by 1 message if necessary(QPID-6088).    On committing 
the received message, the client takes care to revert the prefetch level back 
to the session's default value.   However, if the application calls rollback, 
the client fails to account for the increased credit, so when the application 
calls #receive again, the prefetch window will be opened further.

If the downstream application experiences some kind of failure meaning it rolls 
back every message, this could potentially lead to the prefetch buffer growing 
large.  If the queue is deep, this could cause excessive network utilisation 
(as large numbers of messages are continually received/rejected) or excessive 
memory use on the client.

  was:.


> [0-8..0-91] Prefetch accounting incorrect when message rolled back
> ------------------------------------------------------------------
>
>                 Key: QPID-7394
>                 URL: https://issues.apache.org/jira/browse/QPID-7394
>             Project: Qpid
>          Issue Type: Bug
>          Components: Java Client
>            Reporter: Keith Wall
>
> When the application calls {{Session#receive()}}, the client will 
> automatically open the prefetch window by 1 message if necessary(QPID-6088).  
>   On committing the received message, the client takes care to revert the 
> prefetch level back to the session's default value.   However, if the 
> application calls rollback, the client fails to account for the increased 
> credit, so when the application calls #receive again, the prefetch window 
> will be opened further.
> If the downstream application experiences some kind of failure meaning it 
> rolls back every message, this could potentially lead to the prefetch buffer 
> growing large.  If the queue is deep, this could cause excessive network 
> utilisation (as large numbers of messages are continually received/rejected) 
> or excessive memory use on the client.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to