Login Node Quick Facts
- When you ssh to any NERSC computer system (such as ssh edison.nersc.gov), you are connecting to a "login node."
- The login nodes sit behind a load balancer. New connections are assigned to a login node on a round robin fashion with the exception that if you've connected to a system recently, the load balancer will attempt to put you on the same login node as your previous connection.
- Login nodes are used to edit files, compile codes, and submit job scripts to the batch system to run on the "compute nodes."
- The login nodes are external to the main system so you can log in and work with data and submit jobs when the compute portion of a system is undergoing maintenance.
Process Limits and Appropriate Use of Login Nodes
Most NERSC systems do not have process limits, but please be considerate of other users and do not attempt to run compute-intensive or large-memory jobs on the login nodes. CPU- and memory-intensive applications should be run on the compute nodes and submitted through the batch system. The following applications can be CPU and memory intensive and are popular applications on the login nodes:
Launching any of the above applications for a short time (< 1 hour) on a small dataset should not cause any problems for other users on the login nodes. If you need to run any of these applications for an extended period of time on large datasets, please run them in the batch queues. Our Cray systems supports Python, IDL, Matlab , and NCL.
A few other tips:
- Most NERSC systems have a dedicated xfer queue. If you need to do large number of data transfers to archival storage with hsi or htar, please use this queue to avoid stressing the login nodes.
- Cori has a dedicated interactive queue. You can get up to 20 nodes (Haswell or KNL) for up to four hours within 6 minutes. This is a good resource for debugging scripts or code at scale.
- As a general guideline, a process on the login node should never use more than half the processor cores or a quarter of the memory available on a login node.
- The 'watch' command should be used sparingly.
- Long running user applications should never be run on the login nodes.
- Avoid long compiles with make -j [n] where n is greater than half the number of cores on a login node
- You can determine the number of cores with the command
% cat /proc/cpuinfo | grep processor
NERSC reserves the right to kill processes on the login nodes if responsiveness is being impacted.
Automatic Logout of Idle Sessions
Most NERSC systems will automatically log you off if your session is idle for 48 hours. Idle sessions on PDSF and Genepool are never terminated.