November 21, 2023
All Scientific Computing (SciC) Linux Clusters workloads must be submitted to the job scheduler. Once you are logged on to erisone.partners.org over SSH you will be on remote computer named "eris1n2" (or similar internal name). These are the login nodes.
Please read through all necessary pages under http://rc.partners.org/kbase/High_Performance_Computing, particularly the page for first time users and the page of instructions for job submission
Use of the Login Node
When logged on to one of the login nodes (eris1n2 etc.) please restrict usage to these general areas:
- Submitting batch jobs to the cluster
- using the "bsub" command
- Editing scripts
- Various command line ("nano", "vi", "emacs") and graphical ("nedit", "vim" ) editors are available.
- Copying data to/from the cluster
- "rsync", "scp", "unison"
- File operations
- "mv", "rename", "rm", "find"
- File compression (use nice)
- File compression with "gzip" or "zip" can run on the login nodes, but please prefix the zip command with "nice" eg:
nice gzip my_file
- Getting application help/usage
- All software on the cluster is accessible on the login nodes, however please run the applications through the scheduler
Why does my account have "Restricted Resources" on the login nodes
If you see a message "This account has restricted resources on the login node" then a temporary restriction has been placed on the account. Please contact Scientific Computing for assistance with submitting batch jobs to the cluster.
The login nodes are a shared resource and have many people using them at any one time. Running SciC Linux Clusters applications on the login node can cause other users sessions to be slow or non-responsive. If your account is causing this, the Scientific Computing staff may place a temporary limit on your account (setting "ulimit restrictions") for the login node only. The restriction will be lifted once we've identified the cause - often this can be rectified by changing the bsub command line used.
The temporary "ulimit" restriction will not interfere with most of the accepted uses of the login node described above so work on the cluster can continue as normal. The ulimit will cause long running CPU or memory intensive processes on the login node to exit.