synapse-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Asankha C. Perera (JIRA)" <j...@apache.org>
Subject [jira] Commented: (SYNAPSE-287) An Invalid configuration can lead to lockup
Date Wed, 07 May 2008 09:23:56 GMT

    [ https://issues.apache.org/jira/browse/SYNAPSE-287?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12594835#action_12594835
] 

Asankha C. Perera commented on SYNAPSE-287:
-------------------------------------------

Thanks for the configuration, Can I also know how and where the "http://10.33.31.108:8180/GipMock/"
service is created and hosted as well, so that I can reproduce this scenario.

However, a strong point to note is that it seems like you are running Synapse on a Windows
box! On Windows, the non-blocking transport implementation falls back to using in-memory buffered
streams to convert from NIO channels to streams. Thus I suspect that you may be running into
a low memory issue - is this visible from your JMX monitoring as well? Whats the size of the
heap memory specified through the "-Xms" and "-Xmx" parameters at startup?

I strongly suggest switching the Synapse instance to Linux for load testing, where the native
Pipe's are used for the channel to stream bridging

> An Invalid configuration can lead to lockup
> -------------------------------------------
>
>                 Key: SYNAPSE-287
>                 URL: https://issues.apache.org/jira/browse/SYNAPSE-287
>             Project: Synapse
>          Issue Type: Bug
>          Components: Transports
>    Affects Versions: 1.1.1
>         Environment: Windows XP SP2, JDK SUN 1.5.0_15
> Ubuntu Server 6.10, JDK SUN 1.5.0_15
>            Reporter: Nicolas KOHUT
>            Assignee: Asankha C. Perera
>            Priority: Minor
>         Attachments: 10000_gros_message_KO_CPU.JPG, read_time_out.JPG, synapse.xml, test
de perf.jmx
>
>
> Synapse has some networks problems.
> When it is stressed, is freeze synapse in transport http.
> This behavior is obtained by influencing 2 parameters:
> - the number of competitors access (number of thread).
> - the size of messages sent.
>  
> More generally, the higher the speed is high (combination of 2 parameters above) synapse
freezes more quickly.
> A simple test case (using JMeter):
> - Install distribution synapse-1.1.1 default.
> - Setting up a proxy with sequences in and out empty.
> - Launch a thread (1 single thread is sufficient ...) which calls the proxy with a message
6Ko infinite loop. The thread sits 100 ms between each call.
> Synapse must freeze (random) after about 2 000 calls (5 max 000).
> The freeze takes place at a rate of 7 messages (6Ko) per second ...
>  
> The only configuration synapse that I could find is the file in the folder synapse.properties
Conf of synapse.
> This file can affect a pool of thread used by synapse. Increase the various parameters
of the pool has no impact and does not correct the problem.
> Listed JVM, jconsole indicates that there is no leakage and no memory thread in deadlock.
> Finally, the synapse approved http requests with nio-http (I do not know more).
> Conclusion:
> The simple test presented above is surprising. Synapse seems not to hold a charge very
light.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


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


Mime
View raw message