This response is very much in the "new direction or place to search" category.
Sagemath's build process has been undergoing quite some changes. It used to be the case that sage-the-distribution kept virtually everything in-house, so that an install on shared folder would work great on a cluster with many (identical) nodes. The build system now has much more preference for relying on system-supplied components. As long as all nodes are still identical *and have the same system-supplied components in place*, it should still work. Sagemath is now quite viably buildable in conda, where conda is used to supply many components. So if a plain install isn't working on a cluster (or looks potentially problematic), it could be very instructive to read up on how conda deals with such situations -- that might well be applicable to sagemath (built in conda) now as well. On Thursday, 10 August 2023 at 15:13:54 UTC+2 Andrew wrote: > I've done some research, googled around, searched though ask sage, looked > at some of the thematic tutorials and have finally come to this google > group (ask sage, in all fairness, never approved this post, deleted me and > it, and no idea why and it was from my corporate email address? But ask a > question on the sage page leads there...). > > Can anyone point me towards some documentation, how-tos, or git gist in > and around installing sage on a multi-node HPC cluster? We have a job > manager (PBS family, so like PBS Pro or Torque). Multiple compute nodes, > use modules for the environmental variables > > That said, it is not clear to me about the best approach to laying this > down. Some software lends itself to a shared folder install (though that is > often built from source or installed on a master node), others with a local > install on each node. Maybe it is some kind of mental block, but have > installed others without issue and sliced and diced this question before. > > Just curious about the best approach to install sage (I did find this bit > for mpi support, sage -pip install mpi4py but I can't imagine there isn't > more information that I am somehow just not searching right for) > > Even a new direction or place to search would be a great answer. > > Thanks, > > A > -- You received this message because you are subscribed to the Google Groups "sage-support" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-support+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/sage-support/fdbdbc0e-0997-4bed-8325-2eb6da0b89b9n%40googlegroups.com.
