|Posted on Sunday, September 27, 2015 - 7:23 pm: |
We are running WebMO 18.104.22.168 with nwchem-6.3. If we run NWChem on a single processor, it works fine. If we try to run a parallel job, it fails with a message that it can't find the binary nwchem. I had thought I had compiled nwchem with parallel support, and there is a parallel.x file in the same binary directory as nwchem. Does a parameter for the PE environment need to be set?
Post Number: 473
|Posted on Monday, September 28, 2015 - 1:57 pm: |
You may need to customize run_nwchem.cgi to handle parallel jobs support. There are MANY ways to compile NWChem for parallel support, some that use MPI, others that don't. As such, there is no simple way to make this work "out of box".
|Posted on Tuesday, October 06, 2015 - 1:43 pm: |
Is this still an issue? I have compiled and installed the parallel version of NWChem 6.3 on several plain vanilla CentOS 6 systems - after much experimentation. I put a link to a text file "How To Compile and Install NWChem-6.3 on a CentOS 6.6 System" on the Univ of Alaska WebMO homepage. Note that the process is version specific: what works with CentOS 6.6 might not work with 6.5, and very probably not with CentOS 7. Same goes for NWChem. There are many versions of NWChem 6.3 - the one I have used is 2014-01-28. http://chem.uaf.edu/facilities/WebMO/
|Posted on Wednesday, November 04, 2015 - 4:16 am: |
Hi again, NWChem 6.6 is now available, and compiling is somewhat easier. The online documentation under "Compiling NWChem" has exact procedures for several Linux distros, Mac OS and Windows. I used the "NWChem 6.6 on RedHat 6" successfully on several Intel machines and a VMware Workstation virtual machine.
|Posted on Friday, November 13, 2015 - 8:09 pm: |
Now trying to run NWChem 6.6 on a CentOS 7 remote server, however the WebMO (v 15) remote user cannot run NWC. "Cannot execute /usr/local/nwchem/bin/nwchem: No such file or directory". In fact, the file is there and has "775" permissions, just like all other systems here. Gaussian09 and Mopac2012 are installed on this machine and these jobs run normally when invoked by the webmo head node. Also, NWC jobs run normally when the webmo user logs on locally, ala the test scenario under NWChem section of Support & Tutorials. The only thing I can see is that when the "who" command is entered on the Centos 7 system, the webmo remote user is shown as "?", while on our CentOS 6 systems here, "who" shows the remote user as "+ tty1" So there is something wrong with the remote webmo user account in CentOS 7.
Has anyone used WebMO to run NWC 6.6 jobs on a CentOS 7 system remote server?
Post Number: 482
|Posted on Friday, November 13, 2015 - 10:24 pm: |
This may be a long shot, but I believe Centos 7 has seLINUX enabled by default. seLINUX enables another, orthogonal, layer of security that is very hard to understand and debug. You might try disabling seLINUX and seeing if that resolves the issue. (This is DEFINITELY required if running WebMO on the server -- not sure about remote servers.)
|Posted on Saturday, November 14, 2015 - 8:20 pm: |
I disabled selinux on this machine ("sestatus" command gave "disabled" response), but I'm still seeing the ? user, and NWchem jobs still fail.
I'll let you know if I find out anything more.