activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From TOPPER_HARLEY <TREVOR.HIGG...@ERICSSON.COM>
Subject Re: AMQ 100% of CPU in RCP application
Date Fri, 21 Dec 2007 10:38:16 GMT

Hi,
OK, that does sound reasonable, but if that is the case, then how can you
distinguish between just wait time and actual CPU time  in the profiler
output? 


Hiram Chirino wrote:
> 
> Hi I don't think your interpreting the profiler correctly.
> 
> It is expected that most of the time that thread will be spent on the
> SocketInputStream.socketRead0() call.  But that's just wall time not
> CPU time.  The method is blocking/idle waiting for IO from the socket.
> 
> Regards,
> Hiram
> 
> 
> On Dec 19, 2007 9:58 AM, TOPPER_HARLEY <TREVOR.HIGGINS@ericsson.com>
> wrote:
>>
>> Hi,
>> Its still the same issue in 5.0 by the looks of things:
>>
>> http://www.nabble.com/file/p14418496/AMQ5.0_dump.html AMQ5.0_dump.html
>>
>> 100% is being spent in readInt()
>>
>> Would you expect alot of time to be spent in this method in the first
>> place
>> or is this completely unexpected? Has you/anyone ever seen seen this
>> before?
>> Its always the case on my RCP application so it tried to reproduce it
>> with
>> simple MDP consumer using spring. In this simple test, I was sending 1
>> message every second and when receiving it with MDP, I had an empty for
>> loop
>> 1 million times to try to chew up CPU. On reading the profile, AMQ was
>> still
>> 93% of the application. This may not be so b
>>
>> Does anyone know what to try from here?
>> Thanks
>> /Tom
>>
>>
>>
>>
>> rajdavies wrote:
>> >
>> > This looks like it might be a bug - but ActiveMQ 5.0 is now available
>> > - please give it a try ;)
>> > On Dec 19, 2007, at 10:57 AM, TOPPER_HARLEY wrote:
>> >
>> >>
>> >> Hi,
>> >> Im running an eclipse RCP application that uses AMQ4.1.1 for
>> >> receiving data
>> >> from temp queues(request/response) and topics (for updates). We are
>> >> having
>> >> trouble with CPU on client side and from using YourKit profiler to
>> >> monitor
>> >> this over about 16 hours I see that AMQ is 100% of the time.
>> >> According to
>> >> the snapshot attached, my methods are completely negligible. Here is
>> >> a quick
>> >> snippet:
>> >>
>> >> <All threads>
>> >> 318,332,979  java.lang.Thread.run()  100 %
>> >>    318,003,121  org.apache.activemq.transport.tcp.TcpTransport.run()
>> >> 100 %
>> >>        317,369,872
>> >> org.apache.activemq.transport.tcp.TcpTransport.readCommand() 100 %
>> >>            317,369,747
>> >> org.apache.activemq.openwire.OpenWireFormat.unmarshal(DataInput) 100 %
>> >>                317,369,747  java.io.DataInputStream.readInt()  100%
>> >>
>> >> Generated by YourKit Java Profiler 7.1 EAP build 3004    19-Dec-2007
>> >> 10:19:15
>> >>
>> >> This seems strange since its an RCP application with glazed lists
>> >> etc. I
>> >> would have thought that my methods (or at least eclipse/glazed lists
>> >> would
>> >> be chewing up the CPU). But from this it seems to be AMQ.
>> >>
>> >> Does anyone have any thoughts on this or am I reading this data
>> >> incorrectly?
>> >>
>> >> Thanks for any input
>> >>
>> >> /Tom
>> >>
>> >>
>> >> http://www.nabble.com/file/p14414902/client_dump.html client_dump.html
>> >>
>> >>
>> >> --
>> >> View this message in context:
>> >>
>> http://www.nabble.com/AMQ-100--of-CPU-in-RCP-application-tp14414902s2354p14414902.html
>> >> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>> >>
>> >
>> >
>> >
>>
>> --
>> View this message in context:
>> http://www.nabble.com/AMQ-100--of-CPU-in-RCP-application-tp14414902s2354p14418496.html
>>
>> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>>
>>
> 
> 
> 
> -- 
> Regards,
> Hiram
> 
> Blog: http://hiramchirino.com
> 
> Open Source SOA
> http://open.iona.com
> 
> 

-- 
View this message in context: http://www.nabble.com/AMQ-100--of-CPU-in-RCP-application-tp14414902s2354p14453315.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.


Mime
View raw message