We have a jukebox for read/writeable optical media on a DU 4.0e system
that is NFS exported to a number of other DU machines. The code that runs
it has a parameter that limits the number of mount daemons allowed to run
at any one time (so process space won't become too full). If that number
is exceeded (by a user trying to mount a volume) a currently mounted
volume needs to be unmounted so its daemon can exit and allow the new one
to start. In doing this, the code waits 30 seconds to see if the volume
to dismount has been accessed -- and will not dismount it if that is
the case.
We were finding that this behavior was not happening, and a new mount
would hang forever once the limit was reached. After spending half a day
with the software support people we found that a user on a machine
that NFS mounted this device always had "dxsysinfo" running, and
apparently it was polling the jukebox volumes more often than 30 seconds.
Stopping dxsysinfo resulted in things behaving as they should. (setting
dxsysinfo to not display disk info did not help).
The question is, does anyone know how to set the polling rate for
dxsysinfo? The man pages do not indicate anything about this. The jukebox
is quite heavily used and we'd like to prevent this interference by
dxsysinfo: about all we can think to do is add a crontab entry for root
on each system to check every 2 minutes or so and kill it. Or perhaps
wrap a script around it that does the same thing. But that seems
a bit rude? I guess I'm hoping there's an undocumented switch that will
let us set the update rate?
Received on Fri Jun 04 1999 - 21:35:16 NZST