todo:
finish document under:
/afs/cern.ch/user/d/dfrodrig/workspace/docs/reports
Add more results to *.sxc :
/afs/cern.ch/user/d/dfrodrig/workspace/docs/testResults
Go get them at:
/afs/cern.ch/user/d/dfrodrig/workspace/GMAN/tests/results
lxb6117, lxb6118 configurations also under ~/workspace/docs/reports/tmp!
Command for running consumer: tail -n 100 -f cons | nc lxb6118 6163 > /tmp/tmp
and not referring to the report:
:: http://lxb6118:8161/admin/topics.jsp
Friday, October 19, 2007
Wednesday, October 17, 2007
Getting this:
Exception in thread "ActiveMQ Transport Server: ssl://localhost:61617" java.lang.OutOfMemoryError: GC overhead limit exceeded
at java.io.ByteArrayOutputStream.(ByteArrayOutputStream.java:60)
at com.sun.net.ssl.internal.ssl.OutputRecord.(OutputRecord.java:56)
at com.sun.net.ssl.internal.ssl.OutputRecord.(OutputRecord.java:66)
at com.sun.net.ssl.internal.ssl.AppOutputStream.(AppOutputStream.java:33)
at com.sun.net.ssl.internal.ssl.SSLSocketImpl.init(SSLSocketImpl.java:519)
at com.sun.net.ssl.internal.ssl.SSLSocketImpl.(SSLSocketImpl.java:433)
at com.sun.net.ssl.internal.ssl.SSLServerSocketImpl.accept(SSLServerSocketImpl.java:255)
at org.apache.activemq.transport.tcp.TcpTransportServer.run(TcpTransportServer.java:153)
at java.lang.Thread.run(Thread.java:619)
From time to time... I'll have to carefully analyse.
right now, I am running the tests veryfing saturation and before saturation. Up to now:
MessageSize | Nmessages | Speed | Saturation | SpeedAfterSaturation | ||
1000B | 16384 | 8822.15 [~1.85s] | 15 Messages | 334M/s[~49s] | ||
8192 | 9102 m/s[~0.9s] | 20 Messages | 334M/s [~24.5s] | |||
32768 | 8936 m/s[~3.6s] | 6 Messages | 334M/s [~98s] |
Tuesday, October 16, 2007
Configuration at lxb6118
James installed Activemq 5.0 (latest release )
configuration files are under
activemq-wrapper.conf (for configuring the activemq wrapping)
activemq.xml (for configuring the activemq broker)
log4j.properties
logs/journal are under /var/lib/activemq
and stopping starting has a dedicated start script :
lxb6117 will keep activemq4.1.1
configuration files are under
/etc/activemq
activemq-wrapper.conf (for configuring the activemq wrapping)
activemq.xml (for configuring the activemq broker)
log4j.properties
logs/journal are under /var/lib/activemq
and stopping starting has a dedicated start script :
service activemq start|stop
lxb6117 will keep activemq4.1.1
Brief notes on remake of the tests
Additional variables to take into account:
Maybe it' snot the same in 5.0
Maybe it' snot the same in 5.0
Be careful (for the future...
> DO NOT USE UNDERSCORES IN YOUR BROKERNAMES.
>
http://mail-archives.apache.org/mod_mbox/activemq-users/200707.mbox/%3C84CFFBCA-88FA-4366-A38D-11720C28010B@gmail.com%3E
>
http://mail-archives.apache.org/mod_mbox/activemq-users/200707.mbox/%3C84CFFBCA-88FA-4366-A38D-11720C28010B@gmail.com%3E
Wednesday, October 10, 2007
meeting with James and Max
Just to have a milestone and work on the coming tasks.
- I presented the work done up to know, briefly explaining what has been done in terms of the scripts using scripting ( based on netcat and wassh.)
- on doubts expressed by me and Max regarding the context of the messaging, James clarified the following points:
a) OGF hasn't still endorsed any messaging structure within the grid: all grid is built on point to point;
b) The messaging system is supposed to be a thin layer. Main idea is to decouple the services from knowing what he has to contact;
c) Some use cases for the moment are:
i) SAM publication: workerNodes/Clients are located anywhere. SAM has a central service, getting all the reports from these test Jobs, but some of them are deployed locally. The idea is that the test results would be sent to the closest broker and this would send the message to the proper Central SAM service.
- I presented the work done up to know, briefly explaining what has been done in terms of the scripts using scripting ( based on netcat and wassh.)
- on doubts expressed by me and Max regarding the context of the messaging, James clarified the following points:
a) OGF hasn't still endorsed any messaging structure within the grid: all grid is built on point to point;
b) The messaging system is supposed to be a thin layer. Main idea is to decouple the services from knowing what he has to contact;
c) Some use cases for the moment are:
i) SAM publication: workerNodes/Clients are located anywhere. SAM has a central service, getting all the reports from these test Jobs, but some of them are deployed locally. The idea is that the test results would be sent to the closest broker and this would send the message to the proper Central SAM service.
Subscribe to:
Posts (Atom)