BRB - cryoEM Workflow

Before contacting SCU about a cluster issue, please work through the troubleshooting steps below. Including this information in your email helps SCU diagnose problems more quickly.

Step 1: Define the Type of Issue

Step 2: Identify the Source of the Issue

• Read the Slurm output carefully and review any error messages
• Check syntax, input files, scripts, MPI settings, thread counts, node counts, and memory
• Determine whether the issue appears to be user error, node-specific, or cluster-wide
• Compare with a similar job that worked previously

Useful Commands

Check Which Node a Job Used

sacct -j slurmID --format=jobid,partition,NodeList

Check the Status of All Nodes

sinfo --long --Node

Execution Issues

Locate the Slurm output file and review the error messages. Common examples include:

• MPI abort — try a different node or partition; if that works, report it to SCU
• Out of memory — reduce MPI ranks or threads
• Illegal instructions — review the submission script carefully

If needed, search for part of the error message online. Relion and cryoSPARC communities often have examples of similar issues.

Performance Issues

• Check whether scratch space is full
• Confirm that jobs are reasonably sized and inputs are appropriately binned
• Test the same job on different nodes or partitions
• Ask lab members whether they are seeing the same behavior

Cluster-wide or node-specific issues should be reported to SCU as soon as identified.

Community Support

The WCMC Cryo-EM Slack forum can be helpful for software-specific and cluster-specific issues. Other users may already have encountered the same error.

When Contacting SCU

Email scu@med.cornell.edu and include:

• The issue type: Execution or Performance
• Relevant Slurm job IDs and the node involved
• The parameters used in test jobs
• The absolute path to the Slurm output file
• Whether other users have reported similar issues
• Screenshots, if relevant