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.

Also note that the group is allocated 100 TB in /n/holyscratch01/jacob_lab, so please be mindful of the disk space you are using and remove or archive files as soon as possible.

Some useful aliases for checking the available nodes and CPUs on the HUCE partitions and disk space 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.