Ticket 6877

Created attachment 9937 [details]
scontrol show config

Hi Adam,

Brian let me know that you were discussing setting up a cluster in the cloud to complement your on-premise cluster.  He said we would send some documentation on setting this up, which I'm happy to provide.  

You can start with the github page for the slurm-gcp project, which provides some details on setting up a stand alone cluster with this tool.
https://github.com/SchedMD/slurm-gcp

Codelabs also has a good tutorial on setting this up:
https://codelabs.developers.google.com/codelabs/hpc-slurm-on-gcp/#0

We have a documentation page on configuring Slurm for multiple clusters:
https://slurm.schedmd.com/multi_cluster.html

If you're going to have separate clusters the more secure way of handling this is to have multiple munge keys.  The accounting documentation has some details on configuring multiple munge keys in the section about infrastructure:
https://slurm.schedmd.com/accounting.html#infrastructure

With multiple munge keys you would need to specify which munge socket was listening for which key.  You can do this with the AuthInfo parameter in the slurmdbd.conf and slurmctld.conf files.
You can read more about AuthInfo in the slurmdbd.conf file in the documentation here:
https://slurm.schedmd.com/accounting.html#slurmdbd-configuration

...and here:
https://slurm.schedmd.com/slurmdbd.conf.html#OPT_AuthInfo

And this is the section on setting AuthInfo in the slurm.conf:
https://slurm.schedmd.com/slurm.conf.html#OPT_AuthInfo

Let me know if you have questions about this.

Thanks,
Ben

Hi Adam,

I wanted to make sure the information I sent you about cloud bursting was helpful.  Let me know if you still need help with this ticket.

Thanks,
Ben

Hi Adam,

I believe the documentation links I sent should provide the information you were looking for.  I'm going to close this ticket, but if you have any follow up questions feel free to let me know and we can reopen the ticket.

Thanks,
Ben