Release Notes

16 Recommendations
FSE process clean-up may be noticed when working with FSE commands. After an FSE command
completes its task, it returns control to the command prompt after a short delay. This is because the
command, after termination, leaves its connection open until the omniORB thread terminates it. omniORB
scans for idle connections at regular intervals. A shorter interval will result in a shorter delay after the FSE
command termination.
To shorten this interval, add the following line to the file omniORB.cfg. This will set the scanning interval
to 1 second:
If omniORB.cfg is not yet present in its directory, copy its template to this directory and add the above
line to it.
Linux specific
omniORB.cfg is located in the directory /etc/opt/fse, and its template is located in the directory
/opt/fse/newconfig.
Windows specific
omniORB.cfg is located in the directory %InstallPath%\etc. Its template can be found in the
directory %InstallPath%\newconfig.
CAUTION: Note that the FSE host does not need to be restarted for changes in omniORB.cfg to take
effect, and that its reconfiguration affects all local FSE processes, not only FSE commands.
If you configure the file omniORB.cfg only to shorten the termination delay for FSE commands as
specified above, you need to consider that omniORB.cfg affects the configuration of the FSE interprocess
communication. For more information, see the FSE User’s Guide, chapter ”Configuring FSE”, section
”Reconfiguring FSE interprocess communication”.
Preventing exhaustion of CORBA sockets
As described in section ”Shortening the termination delay for FSE commands” on page 15, CORBA
technology and its omniORB implementation is used for internal interprocess communication in the FSE
implementation. Each CORBA connection reserves one socket from a pool of available CORBA sockets. If
there are too many concurrent FSE processes running in the FSE implementation, the pool of available
CORBA sockets gets exhausted. This happens when numerous brief FSE operations, especially recalls of
small files, are being triggered for a long period of time.
This problem generally occurs because each FSE process, after termination, leaves its connection open and
the belonging CORBA socket in waiting state, until the omniORB thread terminates it. A solution to the
problem is reconfiguration of omniORB so that it scans for idle connections and terminates them more
frequently.
To prevent CORBA from exhausting all sockets, proceed as follows:
1. Shut down the FSE implementation:
2. On the consolidated FSE system or FSE server, stop the omniORB/CORBA naming service:
3. Remove temporary files created by omniNames:
Linux specific
Windows specific
scanGranularity = 1
fse --stop
omniNames --stop
# rm -r /var/opt/fse/omniorb/*
%InstallPath%\var\omniorb>del *.*