Search Mailing List Archives


Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

[farmshare-discuss] Overnight Matlab jobs in corn

Daniel Becker dub at stanford.edu
Thu Jul 26 12:28:32 PDT 2012


(As a side bonus, it also neatly prevents people from littering the corn machines with zombie 'screen' sessions, as seems to have happened here.)


On Jul 26, 2012, at 12:26 PM, Daniel Becker <dub at stanford.edu> wrote:

> How is "qsub ./job.sh" any more complicated than jumping through hoops with screen, nohup (and why use both?), keeptokens and manually keeping track of exactly what machine your stuff was started on? Doesn't make sense to me...
> 
> D.
> 
> On Jul 26, 2012, at 12:12 PM, Stephen James Wolf <sjwolf at stanford.edu> wrote:
> 
>> The queueing system is for barley and is considerably more complicated than using a screen and nohup. Unless you absolutely need a large amount of cores/ram or the processing speed matters to you using barley is not worth the effort. It appears that you simply weren't SSH'g into the machine you created the screen on. If that is the case then simply SSH into that machine and everything should work fine. 
>> 
>> Stephen
>> 
>> ----- Original Message -----
>> From: "Daniel Becker" <dub at stanford.edu>
>> To: "Open discussion for users of FarmShare" <farmshare-discuss at lists.stanford.edu>
>> Sent: Thursday, July 26, 2012 12:10:33 PM
>> Subject: Re: [farmshare-discuss] Overnight Matlab jobs in corn
>> 
>> Check out <http://farmshare.stanford.edu/>; there's a step-by-step example of how to submit a matlab job at <https://www.stanford.edu/group/farmshare/cgi-bin/wiki/index.php/MATLAB>.
>> 
>> Daniel
>> 
>> 
>> On Jul 26, 2012, at 12:08 PM, David Esteban Manosalvas <demanosalvas at stanford.edu> wrote:
>> 
>>> Thanks, can you point me in the right direction on how to use the queuing system?
>>> 
>>> David
>>> 
>>> ----- Original Message -----
>>> From: "Stephen James Wolf" <sjwolf at stanford.edu>
>>> To: "Open discussion for users of FarmShare" <farmshare-discuss at lists.stanford.edu>
>>> Sent: Thursday, July 26, 2012 12:00:53 PM
>>> Subject: Re: [farmshare-discuss] Overnight Matlab jobs in corn
>>> 
>>> No. The screen only exists on that particular corn machine. When you ssh into just the term "corn" you are connected to the lowest load of any particular machine (of which there are 24 I believe). Everything runs only on that machine, the corn servers do not communicate with each other.
>>> 
>>> On your first SSH you can SSH into corn to get a low load machine. Your command prompt will say "cornxx" where xx is the number (07, 22, etc). To log back in and check your process you need to then ssh into that particular corn machine. 
>>> 
>>> ----- Original Message -----
>>> From: "David Esteban Manosalvas" <demanosalvas at stanford.edu>
>>> To: "Open discussion for users of FarmShare" <farmshare-discuss at lists.stanford.edu>
>>> Sent: Thursday, July 26, 2012 11:56:01 AM
>>> Subject: Re: [farmshare-discuss] Overnight Matlab jobs in corn
>>> 
>>> I am almost sure that I did, but assumning that I did not, shouldn't the MATLAB job keep running and  still appear if I run "top" even if I mistakenly ssh in a different corn???
>>> 
>>> David
>>> 
>>> ----- Original Message -----
>>> From: "Stephen James Wolf" <sjwolf at stanford.edu>
>>> To: "Open discussion for users of FarmShare" <farmshare-discuss at lists.stanford.edu>
>>> Sent: Thursday, July 26, 2012 11:50:15 AM
>>> Subject: Re: [farmshare-discuss] Overnight Matlab jobs in corn
>>> 
>>> Are you sure that you are SSH'g into the same corn machine that you created the screen on? e.g. corn05, corn07, etc and not just corn?
>>> 
>>> Stephen
>>> 
>>> ----- Original Message -----
>>> From: "David Esteban Manosalvas" <demanosalvas at stanford.edu>
>>> To: "Open discussion for users of FarmShare" <farmshare-discuss at lists.stanford.edu>
>>> Sent: Thursday, July 26, 2012 11:47:04 AM
>>> Subject: Re: [farmshare-discuss] Overnight Matlab jobs in corn
>>> 
>>> I am using "nohup" and the comands that I am using to run the code are(after I use PuTTy to ssh):
>>> 
>>> kinit;aklog
>>> screen
>>> keeptoken
>>> (run whatever this tells you too)
>>> nohup \matlab -nojvm -nodisplay < scriptname.m >& driver.log &
>>> ctrl+a+d (to detach from the screen)
>>> exit (to close the ssh session)
>>> 
>>> But when I ssh and do a "screen -ls" the screen is not there anymore and the Matlab job is not there either
>>> 
>>> David
>>> 
>>> 
>>> 
>>> ----- Original Message -----
>>> From: "Stephen James Wolf" <sjwolf at stanford.edu>
>>> To: "Open discussion for users of FarmShare" <farmshare-discuss at lists.stanford.edu>
>>> Sent: Thursday, July 26, 2012 11:36:50 AM
>>> Subject: Re: [farmshare-discuss] Overnight Matlab jobs in corn
>>> 
>>> Can you paste the exact commands you are using?
>>> 
>>> The following should create a matlab process which runs on corn without being kicked
>>> 
>>> In initial terminal:
>>> kinit;aklog 
>>> (insert your password into the prompt)
>>> screen -S matlabtest
>>> 
>>> (new screen pops up)
>>> keeptoken
>>> (run whatever this tells you too)
>>> nohup matlab -nosplash -nodisplay < yourprogram.m &
>>> 
>>> now press ctrl+A+D to detach the screen and close your SSH window.
>>> 
>>> Also make sure to remember which corn box you have remoted into and remote into the same box.
>>> 
>>> Stephen Wolf
>>> 
>>> ----- Original Message -----
>>> From: "David Esteban Manosalvas" <demanosalvas at stanford.edu>
>>> To: farmshare-discuss at lists.stanford.edu, "Jason Scott Bishop" <bishopj at stanford.edu>
>>> Sent: Thursday, July 26, 2012 11:31:03 AM
>>> Subject: [farmshare-discuss] Overnight Matlab jobs in corn
>>> 
>>> All, 
>>> 
>>> I have been trying to run some MATLAB programs in corn, I have been using "nohub" to run it in the backgroud, and "screen" to be able to close my ssh and keep them running.  I have also been using "keeptoken" in the screen to make sure that if the job takes longer than 25 hours, AFS wont kick me out, but still after doing all this, when I close my ssh all my MATLAB jobs are killed.
>>> 
>>> Does anyone have any suggestion on what to do about this??
>>> 
>>> Thanks,
>>> 
>>> David Manosalvas
>>> Aeronautics and Astronautics
>>> _______________________________________________
>>> farmshare-discuss mailing list
>>> farmshare-discuss at lists.stanford.edu
>>> https://mailman.stanford.edu/mailman/listinfo/farmshare-discuss
>>> _______________________________________________
>>> farmshare-discuss mailing list
>>> farmshare-discuss at lists.stanford.edu
>>> https://mailman.stanford.edu/mailman/listinfo/farmshare-discuss
>>> _______________________________________________
>>> farmshare-discuss mailing list
>>> farmshare-discuss at lists.stanford.edu
>>> https://mailman.stanford.edu/mailman/listinfo/farmshare-discuss
>>> _______________________________________________
>>> farmshare-discuss mailing list
>>> farmshare-discuss at lists.stanford.edu
>>> https://mailman.stanford.edu/mailman/listinfo/farmshare-discuss
>>> _______________________________________________
>>> farmshare-discuss mailing list
>>> farmshare-discuss at lists.stanford.edu
>>> https://mailman.stanford.edu/mailman/listinfo/farmshare-discuss
>>> _______________________________________________
>>> farmshare-discuss mailing list
>>> farmshare-discuss at lists.stanford.edu
>>> https://mailman.stanford.edu/mailman/listinfo/farmshare-discuss
>>> _______________________________________________
>>> farmshare-discuss mailing list
>>> farmshare-discuss at lists.stanford.edu
>>> https://mailman.stanford.edu/mailman/listinfo/farmshare-discuss
>> 
>> 
>> _______________________________________________
>> farmshare-discuss mailing list
>> farmshare-discuss at lists.stanford.edu
>> https://mailman.stanford.edu/mailman/listinfo/farmshare-discuss
>> _______________________________________________
>> farmshare-discuss mailing list
>> farmshare-discuss at lists.stanford.edu
>> https://mailman.stanford.edu/mailman/listinfo/farmshare-discuss
> 
> _______________________________________________
> farmshare-discuss mailing list
> farmshare-discuss at lists.stanford.edu
> https://mailman.stanford.edu/mailman/listinfo/farmshare-discuss

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4379 bytes
Desc: not available
URL: <http://mailman.stanford.edu/pipermail/farmshare-discuss/attachments/20120726/0af64039/attachment.p7s>


More information about the farmshare-discuss mailing list