distributed simulation: ksh problem

hi all,

I've set up a cluster of computers for distributed simulations. Both
cdsDPSetupChk and adptest report no errors.
Still, I can't run distributed simulations. The jobs are submitted but
are immediately shown as "ABORTED" in the job viewer. To track down the
problem, I created a wrapper script for rsh that logs all command
lines. In it, I see

rsh remote ksh -s -- -clientPlat=lnx86 -clientHost=local
-cdsPath=/inst/ICOA5251 tools/bin/cdsserv 20 0 local 58471 local 0 20 1
1 1 0 72 1 1 1.4 1.0

rsh remote ksh -s -- -clientPlat=lnx86 -clientHost=local
-cdsPath=/inst/ICOA5251 tools/bin/cdsserv -version

and that's where I have to give up. I haven't worked much mich ksh so
far and don't understand the commands. I suspect that they are the
actual problem since they never complete when I run them locally, but
start a shell.
So my question is: Are these command lines correct? And if yes, what do
they try to do? Is it a way to run
/inst/ICOA5251/tools/bin/cdsserv -version? If so, this command works
typed directly, but not via the ksh command above...

And do I need a special version of ksh? I've tried both "sh (AT&T Labs
Research) 1993-12-28 q" and "@(#)PD KSH v5.2.14 99/07/13.2", both
without success.

Blondie

 

I just couldn't give up like that and went on to fight with environment
variables, firewalls, ...

Finally I managed to get cadence to start spectre on a remote host. The
simulation fails however with

Error found by spectre during Sweep analysis `sweep'.
sweep: Stop limit must not equal start limit.

Analysis `sweep' terminated prematurely due to error.

The error apparently is caused by this fragment in input.scs:

sweep sweep param=Iteration start=4 stop=4 step=1 {
tran tran stop=10n errpreset=conservative noiseseed=Iteration
noisefmax=5G \
noisescale=2 write="spectre.ic" writefinal="spectre.fc" \
annotate=status maxiters=5.
}


I'm trying to run a parametric transient noise analysis from within
ADE.

cadence versions: ICOA 5.2.51, MMSIM 6.1, both latest ISR.

Any ideas?

Blondie

 

One important note: The error is not caused by the parametric analysis,
but by the ?tranNoiseMultiRuns and ?noiseruns
options to the transient noise analysis. So far all simulation types
except for multi-run transient noise analysis worked in my setup.

Blondie