qpid-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "rajith attapattu" <rajit...@gmail.com>
Subject Re: Review Request: Sending of completions should be independent of sending message acks
Date Fri, 18 Nov 2011 03:53:28 GMT


> On 2011-11-17 22:50:45, Robbie Gemmell wrote:
> > Also forgot to mention earlier, I think that you will need to consider clearing
the contents of your completions list during failover. When Keith and I made the changes we
did, we had to update the patch to ensure that it didnt try completing messages after failover
that had been received from the old broker and so would now be considered to have invalid
command IDs based on the tracking the Session does.

correct ! -- this is something that I've missed. Thanks again for the feedback.
Hope to repost the patch with the changes we've discussed late friday or monday morning.


- rajith


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/2853/#review3330
-----------------------------------------------------------


On 2011-11-16 18:31:12, rajith attapattu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/2853/
> -----------------------------------------------------------
> 
> (Updated 2011-11-16 18:31:12)
> 
> 
> Review request for qpid, Gordon Sim, Robbie Gemmell, Weston Price, Keith Wall, and Oleksandr
Rudyy.
> 
> 
> Summary
> -------
> 
> This patch is for both QPID-3612 & QPID-3613. Please note QPID-3602 is the main JIRA
for credit related issues.
> 
> Message completions are now handled independent of the Message acks.
> 
> 1. The consumer now keeps track of credits on a per subscription basis using the RangeSet
"completions" and sends completions,
>    if the unCompletedCount >= _capacity/2
>    For capacity = 0 or capacity = 1, we use the flag "sendCompleteForEveryMsg" to send
a completion for each message.
> 
> 2. Cleanedup the acking process in AMQSession_0_10.java to make acking independent of
completions.
>    2.1 For AUTO_ACK we send an ack after every message.
>    2.1 For DUPS_OK we ack if,
>           maxAckDelay <= 0 OR
>           unackedCount >= ackBatchingThreshold, where ackBatchingThreshold == prefetch/2
> 
> 
> This addresses bug QPID-3602.
>     https://issues.apache.org/jira/browse/QPID-3602
> 
> 
> Diffs
> -----
> 
>   http://svn.apache.org/repos/asf/qpid/trunk/qpid/java/client/src/main/java/org/apache/qpid/client/AMQSession_0_10.java
1202779 
>   http://svn.apache.org/repos/asf/qpid/trunk/qpid/java/client/src/main/java/org/apache/qpid/client/BasicMessageConsumer_0_10.java
1202779 
>   http://svn.apache.org/repos/asf/qpid/trunk/qpid/java/systests/src/main/java/org/apache/qpid/client/prefetch/PrefetchBehaviourTest.java
1202779 
> 
> Diff: https://reviews.apache.org/r/2853/diff
> 
> 
> Testing
> -------
> 
> Added two basic test cases (transacted and client-ack modes) to ensure completions are
sent in time to ensure credits don't dry up.
> 
> 
> Thanks,
> 
> rajith
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message