If you recently logged in, you probably already noticed it: now, when you connect to Sherlock, you can see some new status information displayed right before your command prompt.

New status information at login

At login, you’ll know get some some cluster-level data, such as:

  • Sherlock’s status, as reported on the Status Dashboard
  • the current utilization level of the whole cluster, in terms of allocated CPUs

and since we all like to know more about our own usage, you’ll also find:

  • a quick recap of your current jobs on Sherlock (number of jobs running and pending, and corresponding amount of CPU cores),
  • an estimate of the average wait time for a standard job (1 CPU, 2 hours) submitted to the normal partition.

This should hopefully help get a better sense of Sherlock’s utilization levels when you connect.

Show me!

In practice, it looks like this:


In case one of the different stats is not available, or takes too long to get, it will just get skipped and n/a will be displayed in its place. Nobody likes waiting for a shell to appear!

If you don’t care for these stats, you can easily disable them by creating a ~/.sh_noinfo file in your $HOME directory:

$ touch ~/.sh_noinfo

and the status information will be gone the next time you connect.

Individual stats

If you’d like to get individual pieces of information while you’re already connected, to see if they changed or how they evolve, you can use the following

sh_status will display the cluster’s services availability and ongoing incidents, as reported on the Status Dashboard

$ sh_status
OPERATIONAL | uptime : 99.991%

sh_usage reports the overall utilization of the cluster:

$ sh_usage
82.07% | use/tot: 20,160/24,564 cores

sh_jobs will show some stats about your jobs:

$ sh_jobs
1 RUNNING (4 cores), 1 PENDING (4 cores)

sh_jobwait will display the estimated wait time in queue for a typical job that would be submitted now, in the normal partition:

$ sh_jobwait
8 hours and 23 minutes in normal

You can specify a different partition in argument to get the estimate for that partition, as in sh_jobwait owners, for instance.

As usual, if you have any question, please don’t hesitate to reach out to us at srcc-support@stanford.edu