[omniORB] Back on the list again...
Sai-Lai Lo
S.Lo@uk.research.att.com
07 Apr 2000 11:18:04 +0100
Helmut,
Can you tell me the particulars of your platform, (compiler, c library
version, os version)?
When you say "shutdown", how do you go about doing that? Do you use the BOA
shutdown and destroy?
2.8.0 has fixed a well hidden race condition in the idle connection code,
you may want to upgrade to it at some stage.
Sai-Lai
>>>>> Helmut Swaczinna writes:
> Hi everybody,
> before im going to talk about our current problem let me say some positiv
> words:
> Our real complex program system based on omniORB 2.7.1 under Linux works
> now in a real-life, commercial environement since several months. It runs
> nigth and day, week for week without any crash or the need to shutdown.
> Now the "but":
> When it is necessary to shutdown the omniORB-processes (e.g. for
> configuration
> tasks) after *some weeks* the system shows a very strange behaviour. The
> systemload increases to an exorbitant value. It is nearly impossible to
> execute a command on the console. Shuting down the processes takes up to an
> hour on a quit fast machine. Normaly it takes a few seconds. The heavy
> systemload is caused by a *very* high traffic rate on the loopback-device.
> There are some hundred million packets delivered during shutdown. During
> normal operation there are some million per week. This traffic must be located
> in the omniORB-library. There are no extra method-invokations or upcalls in
> the user-code.
> The system consists of 12 omniORB-processes each with about 10 threads
> running on one machine. All are combined client/server-processes. The
> processes are startet with -ORBinConScanPeriod 0 ans -ORBoutConScanPeriod 0.
> Scanning for idle-connections makes the system get instable.
> Can somebody explain this strange behaviour to me?
--
Sai-Lai Lo S.Lo@uk.research.att.com
AT&T Laboratories Cambridge WWW: http://www.uk.research.att.com
24a Trumpington Street Tel: +44 1223 343000
Cambridge CB2 1QA Fax: +44 1223 313542
ENGLAND