Cluster Rules

Amy Keating, a BE faculty member, administers coyote. She has four rules for the cluster:

1. ALL jobs must be submitted through the queuing system. It is not ok to directly run jobs on any of the nodes (other than the designated interactive nodes ethelbert and supergenius)
2. No jobs should ever be run on the head node.
3. Do not submit more than 500 jobs at a time to most queues (1000 jobs in the "quick" queue; 10 jobs in the ultra queue)
4. Comply promptly with all requests from me, my postdoc Vladimir Potapov, or our system administrators (Greg, at TechSquare), e.g. if asked to kill jobs or to change how you are using the cluster.

Things to note

• You can ssh to coyote only from MIT IP addresses. This means you either need to be on campus or be connected via the MIT VPN. (This was to mitigate security attacks on the cluster.)
• The default shell on coyote is tcsh. To change to bash, issue: chsh -p /bin/bash my_username. You might need to use usermod rather than chsh.

Getting started

• Email Greg (a.k.a. "[greg]") (bio-admin@techsquare.com). Greg will request confirmation that you're in the lab, so it's best if you immediately cc a lab member that Greg already knows.

• Consider adding coyote to you ~/.ssh/config file.
• Make sure you understand the difference between the head node and the queues before you start doing any work.