Cannon cluster etiquette

Be mindful when submitting jobs

Recall that the huce_intel and huce_cascade partitions have no time or memory limits. This can be a both blessing and a curse at the same time. While it is possible to submit many long simulations to these queues, we discourage this in practice. If you submit a large number of jobs you may end up blocking other users' jobs from running.

Our rule of thumb is that if you need to use a large amount of computational resources to please alert the group by sending a quick message to the #fasrc-cluster channel of the Harvard-ACMG Slack Workspace.

Some useful aliases for checking the available nodes and CPUs on the HUCE partitions can be found in our sample .bashrc file:

# List all jobs running by user (e.g. "sq -u $USER") or
# partition (e.g. "sq -p huce_cascade")
alias sq="squeue --format='%.10i %.10P %.10j %.10u %.10a %.2t %.10M %.6D %.6C %R'"
alias squ="sq -u $USER"

# View information about Slurm nodes and partitions
alias si="sinfo format='%.11P %.20N %.11T %.6D %.4c %C'"

# List number of nodes and CPUs in a given partition
# (A=available, I=idle, O=other, T=total)
alias shinfo="sinfo --format='%.11P %.20N %.11T %.6D %.4c %C' -p shared"
alias tinfo="sinfo --format='%.11P %.20N %.11T %.6D %.4c %C' -p test"
alias hiinfo="sinfo --format='%.11P %.20N %.11T %.6D %.4c %C' -p huce_intel"
alias hcinfo="sinfo --format='%.11P %.20N %.11T %.6D %.4c %C' -p huce_cascade"
alias hbinfo="sinfo --format='%.11P %.20N %.11T %.6D %.4c %C' -p huce_bigmem"

# Check my fairshare score
alias myfs="sshare -u $USER -U"

# Check Jacob group fairshare score
alias jacobfs="sshare -A jacob_lab -a"

More information

Please see the FASRC Cluster Customs and Responsibilities Guide for more best practices for using Cannon.