WebMO - Computational chemistry on the WWW
Recent news

WebMO 15.0 is now available for free download!

WebMO 15.0 Pro and Enterprise have a variety of additional features and is available for purchase.

The WebMO app for iPad/iPhone is now available on the App Store.

August 23, 2019

NWChem 6.1.1 running - but leaves >1... Log Out | Topics | Search
Moderators | Edit Profile

WebMO Support Forum » Running WebMO Jobs » Other » NWChem 6.1.1 running - but leaves >12 GB of files « Previous Next »

Author Message
John Keller
Unregistered guest
Posted on Saturday, February 02, 2013 - 1:43 am:   

I got the parallel version of NWChem 6.1.1 compiled and installed on CentOS 6.3 (Redhat derivative) using the openmpi 1.5.4 that comes with CentOS. However, each job is leaving lots of big input.aoints.0n files in the /scratch/webmo-14480/username/jobnumber folder. They have -rw--r--r-- permissions, whereas the usual output files such as output.out are -rw-rw-r--.
JR Schmidt
Moderator
Username: Schmidt

Post Number: 330
Registered: 11-2006
Posted on Saturday, February 02, 2013 - 10:13 am:   

What queuing system are you using?

These should be deleted at the END of the job. Are these being left for successful jobs, or for crashed jobs?
John Keller
Unregistered guest
Posted on Sunday, February 03, 2013 - 1:44 am:   

I am using the WebMO queuing system. Its not quite as bad as I originally thought: only one job had saved all those files (but they were moved from the compute node to the user's folder on the webmo server! I deleted them.)
Still: for parallel jobs an approximately-1-MB input.gridpts.n file is saved and moved to the user's folder for each processor requested in the NWChem job. See the attached txt file.
text/plainFile listing for successful NWChem 16-proc job
job1677filelist.txt (2.4 k)
JR Schmidt
Moderator
Username: Schmidt

Post Number: 331
Registered: 11-2006
Posted on Monday, February 04, 2013 - 10:32 am:   

Interesting. I will try adding a "scratch_dir" specification to the NWChem job templates to direct these files elsewhere.
John Keller
Unregistered guest
Posted on Saturday, March 02, 2013 - 2:37 am:   

OK, that works. IF, I edit the input file by adding the line

scratch_dir /scratch/webmo-3164/NNNN

where NNNN = the webmo job number. This is a short-term fix for this problem which affects almost all NWChem 6.1.1 jobs. Ideally a variable name should go there. I will bet this will require changes in run_nwchem.cgi.
John Keller
Unregistered guest
Posted on Saturday, March 02, 2013 - 2:46 am:   

Also: This also presupposes the user has just run a job, so they have a good indication what the next job number will be (unless someone else is logged on...)
JR Schmidt
Moderator
Username: Schmidt

Post Number: 338
Registered: 11-2006
Posted on Saturday, March 02, 2013 - 9:32 am:   

If you update to the newest release this is taken care of automatically.
John Keller
Unregistered guest
Posted on Monday, March 04, 2013 - 3:47 am:   

If this problem also occured with NWChem 5.n or 6.0 parallel versions, then perhaps it should be considered a WebMO bug. On the other hand, if this problem occurs only with the newest verson of NWChem, v 6.1, then that is just an unfortunate software change that does warrant our purchasing a WebMO update. Am I off-base?
JR Schmidt
Moderator
Username: Schmidt

Post Number: 339
Registered: 11-2006
Posted on Wednesday, March 06, 2013 - 12:37 pm:   

In general, our policy is that we cannot back port enhancements and bug fixes to older, unsupported version of WebMO (this is simply unfeasible!). In general, I would recommend upgrading to the most recent supported version, which incorporates these (and many other) features and enhancements.

That said, in this case, you can probably handle the back port yourself: If you download the most recent build of WebMO 13 "free" and look in run_nwchem.cgi, you will see the relevant changes to redirect the scratch files. You should be able to port these changes into your older WebMO version to similar effect.

Hope that helps!
John Keller
Unregistered guest
Posted on Tuesday, November 18, 2014 - 11:53 pm:   

Still have a problem running NWChem 6.3 where, if it fails, it leaves processes going on CPU threads (NWCHEM). If one issues a "killall -9 nwchem", the processes are removed and the node can again be used by WebMO. Is this an NWChem issue or a WebMO issue, or both?
JR Schmidt
Moderator
Username: Schmidt

Post Number: 434
Registered: 11-2006
Posted on Wednesday, November 19, 2014 - 1:26 pm:   

To be clear:

All is well for successful jobs, but processes linger if the job crashes.

Does this happen if you run jobs via the command line? If so, it is strictly an NWChem issue. If not, I can take a look.

Add Your Message Here
Post:
Username: Posting Information:
This is a public posting area. Enter your username and password if you have an account. Otherwise, enter your full name as your username and leave the password blank. Your e-mail address is optional.
Password:
E-mail:
Options: Post as "Anonymous"
Enable HTML code in message
Automatically activate URLs in message
Action:

Topics | Last Day | Last Week | Tree View | Search | Help/Instructions | Program Credits Administration