[
https://issues.apache.org/jira/browse/QPID-3252?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13031384#comment-13031384
]
[email protected] commented on QPID-3252:
-----------------------------------------------------
bq. On 2011-05-10 14:36:31, Alan Conway wrote:
bq. > Does this affect performance at all?
bq. > Looks ok to me but I'm not up on why we need to flush here.
Performance?
Dang you, Conway! Turns out it does affect the performance of the transient
case - about 10% hit.
I've got a refactored patch that avoids any unnecessary overhead in the path
where messages complete synchronously. I'll upload that now.
- Kenneth
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/701/#review657
-----------------------------------------------------------
On 2011-05-09 21:22:38, Kenneth Giusti wrote:
bq.
bq. -----------------------------------------------------------
bq. This is an automatically generated e-mail. To reply, visit:
bq. https://reviews.apache.org/r/701/
bq. -----------------------------------------------------------
bq.
bq. (Updated 2011-05-09 21:22:38)
bq.
bq.
bq. Review request for qpid, Alan Conway, Gordon Sim, and Kim van der Riet.
bq.
bq.
bq. Summary
bq. -------
bq.
bq. Broker now issues an explicit flush for each received message that has the
sync flag set in the message.transfer command.
bq. Broker also tracks all messages received that do _not_ have the sync flag
set, and will flush these messages on receipt of an execution.sync.
bq.
bq.
bq. This addresses bug QPID-3252.
bq. https://issues.apache.org/jira/browse/QPID-3252
bq.
bq.
bq. Diffs
bq. -----
bq.
bq. /trunk/qpid/cpp/src/qpid/broker/SessionState.h 1101206
bq. /trunk/qpid/cpp/src/qpid/broker/SessionState.cpp 1101206
bq.
bq. Diff: https://reviews.apache.org/r/701/diff
bq.
bq.
bq. Testing
bq. -------
bq.
bq. qpid unit & store unit
bq.
bq.
bq. Thanks,
bq.
bq. Kenneth
bq.
bq.
> Regression: broker no longer explicitly flushes messages to store on
> execution.sync or when sync command bit set.
> -----------------------------------------------------------------------------------------------------------------
>
> Key: QPID-3252
> URL: https://issues.apache.org/jira/browse/QPID-3252
> Project: Qpid
> Issue Type: Bug
> Components: C++ Broker
> Affects Versions: 0.10
> Reporter: Ken Giusti
> Assignee: Ken Giusti
> Fix For: 0.11
>
>
> In the previous release of the broker, an execution.sync command, or a
> message.transfer command with a sync flag would cause the broker to
> explicitly flush the message(s) and pend until the flushes complete. In the
> new asynchronous model, the broker does not initiate a flush under these
> scenarios. Without explicitly flushing, the completion of a
> message.transfer/execution.sync may delay for an unacceptable amount of time
> (ex. 1sec+) when used with store.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project: http://qpid.apache.org
Use/Interact: mailto:[email protected]
