ocean DRMS and wait statements

R

rick

Guest
what determines if "distributed processing is not available."?

ocean> ocnHelp('wait)

PROTOTYPE wait( jobName [jobName2 jobName3 - jobNameN] ) =>
t/nil


DESCRIPTION Postpones processing of a script until the
specified jobs complete. This command is ignored if
distributed processing is not available.
 
rick wrote, on 09/22/10 23:45:
what determines if "distributed processing is not available."?

ocean> ocnHelp('wait)

PROTOTYPE wait( jobName [jobName2 jobName3 - jobNameN] ) =
t/nil


DESCRIPTION Postpones processing of a script until the
specified jobs complete. This command is ignored if
distributed processing is not available.
You have to be using the hostMode() function to tell it is running in
distributed mode (hostMode('distributed)).

Not sure you can use this if you're using the command line mode of distributed
processing (as it may not have a means of knowing when the job finished).

Regards,

Andrew.
 
On Sep 26, 8:06 am, Andrew Beckett <andr...@DcEaLdEeTnEcTe.HcIoSm>
wrote:
rick wrote, on 09/22/10 23:45:

what determines if  "distributed processing is not available."?

ocean>  ocnHelp('wait)

PROTOTYPE       wait( jobName [jobName2 jobName3 - jobNameN] ) =
                 t/nil

DESCRIPTION     Postpones processing of a script until the
                 specified jobs complete. This command is ignored if
                 distributed processing is not available.

You have to be using the hostMode() function to tell it is running in
distributed mode (hostMode('distributed)).

Not sure you can use this if you're using the command line mode of distributed
processing (as it may not have a means of knowing when the job finished).

Regards,

Andrew.
would any of the ipc commands work?
 
rick wrote, on 10/02/10 00:00:
On Sep 26, 8:06 am, Andrew Beckett<andr...@DcEaLdEeTnEcTe.HcIoSm
wrote:
You have to be using the hostMode() function to tell it is running in
distributed mode (hostMode('distributed)).

Not sure you can use this if you're using the command line mode of distributed
processing (as it may not have a means of knowing when the job finished).

Regards,

Andrew.

would any of the ipc commands work?
Not sure what you mean. Please clarify...

Andrew.
 
On Oct 4, 4:39 am, Andrew Beckett <andr...@DcEaLdEeTnEcTe.HcIoSm>
wrote:
rick wrote, on 10/02/10 00:00:

On Sep 26, 8:06 am, Andrew Beckett<andr...@DcEaLdEeTnEcTe.HcIoSm
wrote:
You have to be using the hostMode() function to tell it is running in
distributed mode (hostMode('distributed)).

Not sure you can use this if you're using the command line mode of distributed
processing (as it may not have a means of knowing when the job finished).

Regards,

Andrew.

would any of the ipc commands work?

Not sure what you mean. Please clarify...

Andrew.
Inter Process communication functions like ipcWait
 
rick wrote, on 10/04/10 15:15:
On Oct 4, 4:39 am, Andrew Beckett<andr...@DcEaLdEeTnEcTe.HcIoSm
wrote:
rick wrote, on 10/02/10 00:00:

On Sep 26, 8:06 am, Andrew Beckett<andr...@DcEaLdEeTnEcTe.HcIoSm
wrote:
You have to be using the hostMode() function to tell it is running in
distributed mode (hostMode('distributed)).

Not sure you can use this if you're using the command line mode of distributed
processing (as it may not have a means of knowing when the job finished).

Regards,

Andrew.

would any of the ipc commands work?

Not sure what you mean. Please clarify...

Andrew.

Inter Process communication functions like ipcWait
I know what IPC functions are! I'm not sure how this has anything to do with
launching a simulation using DRMS mode. The problem is that even if there was an
IPC handle for the command line command that was executed to submit the job
(e.g. "myjobSubmit ... spectre ...") then it would presumably have submitted the
job and then returned straightaway. So the IPC job would have finished - no
indication of when it really really finished.

But I'm not sure if that's what you're asking.

Regards,

Andrew.
 
On Oct 4, 9:43 am, Andrew Beckett <andr...@DcEaLdEeTnEcTe.HcIoSm>
wrote:
rick wrote, on 10/04/10 15:15:



On Oct 4, 4:39 am, Andrew Beckett<andr...@DcEaLdEeTnEcTe.HcIoSm
wrote:
rick wrote, on 10/02/10 00:00:

On Sep 26, 8:06 am, Andrew Beckett<andr...@DcEaLdEeTnEcTe.HcIoSm
wrote:
You have to be using the hostMode() function to tell it is running in
distributed mode (hostMode('distributed)).

Not sure you can use this if you're using the command line mode of distributed
processing (as it may not have a means of knowing when the job finished).

Regards,

Andrew.

would any of the ipc commands work?

Not sure what you mean. Please clarify...

Andrew.

Inter Process communication functions like ipcWait

I know what IPC functions are! I'm not sure how this has anything to do with
launching a simulation using DRMS mode. The problem is that even if there was an
IPC handle for the command line command that was executed to submit the job
(e.g. "myjobSubmit ... spectre ...") then it would presumably have submitted the
job and then returned straightaway. So the IPC job would have finished - no
indication of when it really really finished.

But I'm not sure if that's what you're asking.

Regards,

Andrew.
we have a bunch of ocean scripts for corners (multiple jobs) and
include post-processing. The sim
must be complete before the post-processing starts otherwise it errors
out. Since the job count is high,
it would be very desirable to run in parallel or in distributed mode.
We have been using SGE with a
bunch of extra lines of code will submit multiple jobs. SGE and bsub
have built in hooks for distributed
and will honor a wait statement which will halt the script until the
job has completed and then allow it to
advance to the post processing section. We are trying to use RTDA's
NC job scheduler which will dispatch
the jobs quickly and to the "best" box based on various other
factors. Their command is nc run and
does have any ADE/XL/Spectre built in support but the ?drms run
option allows you use any tool
to submit a job. However, in drms mode, waits have been disabled so
the script blows right through the
wait command. Ive tried the script based (Solution ID:11015590) but
it has the same problem. So we
need a mechanism to halt the code until the sim is complete then start
the post-processing. Not sure the
wait function is disabled by name or if there is some bit that gets
flipped. So a custom wait function or
perhaps the ipc commands could be the workaround. I have not been
able to get the ipcxxxx to work
so far. Do you have any ideas?

Hope this helps.

Rick
 

Welcome to EDABoard.com

Sponsor

Back
Top