I've installed and compiled updated versions of NAMD and VMD on Cyrus1. I'll do the same on Quantum2 this weekend if I can. The following binaries can be found in /usr/group on cyrus1:
catdcd charmrun flipbinpdb flipdcd namd2 psfgen vmd (vmd is technically not a binary)
Users can specify the fulll location or modify their paths, as they see fit. Please note that I copied this directory to all the compute nodes.
More details:
Charm++ 6.2: compiled against OpenMPI, in /usr/src/charm-6.2.0/mpi-linux-x86_64-mpicxx/
NAMD 2.7b2: compiled from source using Charm++ 6.2, in /usr/src/NAMD_2.7b2_Source/Linux-x86_64-g++
VMD 1.8.7: /usr/src/vmd-1.8.7, binary downloaded.
catDCD 4.0
Other users who have installed software should feel free to follow this same model, placing source code in the /usr/src directory and executables in the /usr/group directory.
Erik and I met with a consultant from TechSquare Friday afternoon, who helped us get the OpenPBS/Torque queue system working, and more importantly, explained what exactly what was going on. I'd like to thank Erik for his help coordinating this.
We're now ready to start using the queue system. I would appreciate it if you started your next jobs in the queue system. The transition period may be a bit awkward, until all the previously started jobs are finished, but the system does sometimes notice a full (outside the queues) node and avoid running there.
The full instructions to get started cover a lot of material; please read through them at the group computational page: Policies and Queuing System
Please take special note that some nodes on Cyrus1 (n017 through n021) are reserved for non-queuable work (e.g. Accelrys software, if we can't get it working).
If you have any questions about this, you can contact me, or comment on the Wiki.
We have asked local computing consultant Tech Square to examine the queue system on our quantum2 and cyrus1 clusters. There is one hitch, as Erik described in a recent e-mail:
However, there is a slight problem. Around 1:30-2:00 pm tomorrow Friday, April 16, any jobs that are running in cyrus1 or quantum2 may get killed. So please be prepared for it. I'll send another email after we're done setting things up.
More updates to follow.