[
https://issues.apache.org/jira/browse/QPID-6730?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14744212#comment-14744212
]
Jakub Scholz commented on QPID-6730:
------------------------------------
Thanks for this change. I gave it a try and it seems to work nicely - it helps
with the issue with synchronous message producers. When tested with Qpid JMS
client with ~1kB messages, it scaled from 1msg/s with default values up to
somewhere around 400 msg/s with the flush timeout set to 0.1 ms (running on my
old laptop - for comparison asynchronous producer is able to run around 6000
msg/s).
I noticed only two very minor issues - setting the value to 0 results and an
error loop ("[System] error JournalInactive:TplStore couldn't setup next timer
firing: 118.467ms[0ns]"). Also, I would expect that the values usually used
would be rather in milliseconds and not in seconds. Setting the value to 0.001
(aka 1ms) works fine, but it looks strange.
> Make the journal flush timeout configurable in the C++ broker
> -------------------------------------------------------------
>
> Key: QPID-6730
> URL: https://issues.apache.org/jira/browse/QPID-6730
> Project: Qpid
> Issue Type: Improvement
> Components: C++ Broker
> Affects Versions: qpid-cpp-0.34
> Reporter: Michael Cressman
> Assignee: Andrew Stitcher
> Priority: Minor
> Fix For: qpid-cpp-next
>
>
> Linear store does not flush the buffer often enough for synchronous durable
> messages. Make the journal flush timeout configurable so users can tune
> performance by changing this setting.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
