The first and ideal source for queue information and details about the
queues on a cluster system should be your clusters support
staff. They will be able to brief you on the details of particular queues, and
their intended uses, and any limitations. If further information is needed you
can use these commands to divine queue information from the scheduling system
itself.
To view all the queues configured in your cluster system -
Example:
[jdpoisso@umms-amino ~]$ qstat -Q Queue Max Tot Ena Str Que Run Hld Wat Trn Ext T ---------------- --- --- --- --- --- --- --- --- --- --- - casp 0 0 yes yes 0 0 0 0 0 0 E ambyroy 0 0 yes yes 0 0 0 0 0 0 E default 0 891 yes yes 0 891 0 0 0 0 E interactive 0 0 yes yes 0 0 0 0 0 0 E urgent 0 238 yes yes 0 238 0 0 0 0 E [jdpoisso@umms-amino ~]$
From this output we can see that the cluster umms-amino has five
queues configured. It also lists out an overview of the queues, the number of
jobs, the status of the queues, and the how many jobs are in various job
states. With this information we get a basic summation of the activity on the
system, and which queues are heavily in use.
For our purposes, the only important information are the names of the
queues. The other information, though useful, does not tell us any specific
details on what rules the queue uses. To do this, we need to run a
``qstat -Q -f'' on the queues.
Example:
[jdpoisso@umms-amino ~]$ qstat -Q -f default Queue: default queue_type = Execution Priority = 1 total_jobs = 889 state_count = Transit:0 Queued:0 Held:0 Waiting:0 Running:889 Exiting:0 resources_default.nodes = 1 resources_default.walltime = 01:00:00 mtime = 1270501564 resources_assigned.mem = 83886080000b resources_assigned.nodect = 889 enabled = True started = True [jdpoisso@umms-amino ~]$
Using ``qstat -Q -f'' on the
``default'' queue, were
able to see several pertinent details about the queue and the rules it uses.
For instance we can see the ``Priority'' of
the queue, which is a number assigned to all jobs submitted to the queue to
determine when they should run. The higher
``Priority'' of your job, the sooner it will
run. The other morsel of valuable information in are the lines regarding the
default resources. this information you can see that each job submitted
to this queue is given a single node by default, and a walltime of one hour.
Note, these values are just defaults, meaning that your resource requests may
override them. In this queue you will also notice there are no job minimums,
and no job maximums, these will show up in this output if they are present.
Also if there is a set amount of resources limited available to the queue, that
also will be listed. (Draft Note: Construct an example)
Note: Not all queue rules may be listed in the output of
``qstat -Q -f'' especially for aspects like
node restrictions, and whether the queue is limited to specific users. This is
why first and foremost you should consult with your support staff or reference
any cluster specific documentation for details regarding the queues.