From NWChem
Viewed 148 times, With a total of 4 Posts
|
|
|
Clicked A Few Times
Threads 3
Posts 7
|
|
| 4:32:21 PM PDT - Wed, May 2nd 2012 |
|
Installed ECCE from the 64bit 6.3 binaries on a true (non-virtual) linux box running CentOS6. Gateway opens fine, but builder and view trigger the same "Unable to start Applications" message currently being discussed on another thread. I'm waiting to see the solution found for that thread, but thought I should mention the problem occurs with CentOS6.2 (final) as well.
I thought it might be nice to have the solution (once determined) to be posted on the forums, in addition to the end of a multi-page thread.
If there is anything I can do to help troubleshoot that wouldn't be redundant (build from source, etc.), let me know.
Don Berry
BTW, all the requisite packages and libraries are in place and current.
|
|
|
|
|
|
Gets Around
Threads 1
Posts 49
|
|
| 4:52:31 PM PDT - Wed, May 2nd 2012 |
|
Hi Don,
CentOS is basically the "freebie version" of Red Hat. So not too surprising an issue on one would happen on the other. The good news is that I think I'm very close to a fix for this nasty 64-bit OpenGL library issue. I'm doing some test builds and installs now and if it works, then I'll push out new distributions of ECCE tomorrow (Thursday), both binary and source code. So don't worry about any troubleshooting at this point.
Thanks,
Gary
|
|
|
|
|
|
Gets Around
Threads 1
Posts 49
|
|
| 11:07:17 AM PDT - Thu, May 3rd 2012 |
|
Hi Don,
See the last post on the "ECCE 6.3 apps won't start" thread for the latest with 64-bit OpenGL. Give the latest distribution a try and let me know how it works. I see that you have been switching platforms around since you previously posted in regards to Ubuntu. Are you just trying to find a stable platform to run ECCE or do you have a need for different people to run on different platforms? If it's the former, my advice is Debian as a first choice, Mint as a second choice if you get to choose yourself and don't have specific reasons to choose others. That's my experience in developing and testing ECCE over the past few months. But, others should work as well, especially with this latest 64-bit OpenGL fix.
Gary
|
|
|
|
|
|
Clicked A Few Times
Threads 3
Posts 7
|
|
| 7:44:09 PM PDT - Thu, May 3rd 2012 |
|
Hi Gary,
So, I've been exploring ECCE (and nwchem) in a variety of environments, ultimately for several types of users who can't/won't work at the command line.
The new binaries do seem to fix that openGL problem:
(1) your new binaries worked nicely on a non-virtual CentOS6.2 box (64bit). I connected to the box from my Mac laptop tunneled through "ssh -Y". A few tests of all the gateway components and a nwchem job submission didn't give any errors or messages in the console.
(2) I also installed the new 32bit binaries on a virtual box Ubuntu server (i.e. smaller memory footprint, without the full GUI desktop.) FWIW, the screwed-up ecce menu problem only seems to appear under the Ubuntu "Unity" / GNOME desktop; not in the server version with individual gtk packages installed. I never had to remove appmenu-gtk for the Ubuntu server.
My main task now seems to be understanding how to customize the compute server scripts for our aging PBS cluster.
Oh, and as far as requests go... we've replaced g03 with g09 on our departmental clusters. Although features have been added to g09, all previous g03 input files run under g09. is there a quick-n-dirty way to make a basic "g09 tool", even if the builder just creates the same g03 input file, but allows one to submit it to g09 on the compute server? I understand if you wouldn't want to include that in a public distribution until it was full-featured, but I'd be happy to beta test it for you. I did look into the gensub script, but haven't had time to try to patch it for g09 yet.
Finally, I have had zero success figuring out how to import the results from a gaussian calculation into ecce. Back to the manual...
Cheers,
Don Berry
|
|
|
|
|
|
Gets Around
Threads 1
Posts 49
|
|
| 11:15:37 AM PDT - Fri, May 4th 2012 |
|
Thanks for the update Don and confirmation that OpenGL seems to be straightened out now.
As far as compute server registration, it can be a bit tricky. I assume you found our documentation at http://ecce.pnl.gov/using
There is a compute server registration PDF in the "Related Documents" box. Seems like just as valuable as that is looking at the files in $ECCE_HOME/siteconfig/CONFIG-Examples. We have the "ecce -admin" application for registering compute servers, but I woud personally only use that for simple workstations that don't use a batch queue scheduler.
As far as G09, as you may know EMSL is not allowed to run versions of Gaussian since we were put on the Gaussian blacklist many years ago now. Gaussian 98 was actually the last version we had here and it was another ECCE site that helped with the Gaussian 03 registration! I can probably assist you with this in the next couple weeks as far as getting some templates in place to do a basic Gaussian 09 launch (it might be as easy as tweaking the $ECCE_HOME/scripts/gensub script to fool it into launching g09 when it thinks you are running g03 in ECCE). Then we can figure out whether there are other changes to the input and output for the latest Gaussian that ECCE needs to account for in order to work properly. I'm not much worried about "full-feature" support since our Gaussian registratrion is pretty simplistic anyway. I'd just like to verify that ECCE can do setup and output parsing for basic Gaussian 09 jobs before making it more widely available. I know that Gaussian strives for consistency between input and output files between versions because there is so much software that depends on that. I do remember at one point they changed how an important property was output (can't remember if that was 94->98 or 98->03) and we had to make changes for that in ECCE. If you look in the $ECCE_HOME/scripts/parsers directory you can see all the scripts used to parse code output. Looking at the gaussian* files, I see that geometry step (trace) format changed between 94 and 98 and the molecular orbital output must have changed between 98 and 03. Other than that it looks pretty stable. Anyway, we'd love to get basic support for G09 in ECCE and would appreciate your help in doing that.
Gary
|
|
|
AWC's:
2.5.10 MediaWiki - Stand Alone Forum Extension
Forum theme style by: AWC
