I recently set up our HPC environment on Seqera Platform. We are using a cluster with a SLURM scheduler provided by our University. Since we don’t have root access to the cluster, I was wondering how other people deploy the tower-agent in the background continuously or semi-continuously to be able to launch pipelines from Seqera Platform.
Thanks for the help!
One option is to use a terminal multiplexor like tmux or screen to keep it running in the background.
Thanks @jordeu ! That only stays active while I am connected in my shell to the cluster though correct?
No. When you open a new session with a terminal multiplexor that session is a process detached from the parent, the login session, and if you close the login session the detached session keeps running in the background at the login node.
the main headache with screen / tmux sessions usually seems to be when the Sys Admins reboot the cluster or the node(s) you had it running on; they dont come back after reboot and you have to manually start everything again
Thanks @jordeu and @steve! I didn’t know that terminal multiplexor sessions stay open even if I disconnect. I guess the sysadmins might not allow these types of jobs and cancel them though… That seems like the most easy to implement solution for now then I guess!
They’re not really “jobs” in the traditional cluster submission sense. We and many others have used this approach for running the Nextflow head job for years with happy sysadmins and generally good performance. As @steve says - you just have to watch out for cluster reboots (if this happens a lot then there are more complex setups you can look into, but most will require help from a sysadmin).