Queues

A queue is needed when submitting jobs, if no queue is given the SGE submits your job to your default queue. There are several queues available. Only the administrator can create others if needed. To display the available queues use the qstat or qconf command.

vill@shark:~$ qstat -g c

Prints a cluster queue summary

CLUSTER QUEUE                   CQLOAD   USED    RES  AVAIL  TOTAL aoACDS  cdsuE  
--------------------------------------------------------------------------------
12hr-KG.q                         0.16     12      0     60     72      0      0 
2hr-all.q                         0.16      0      0     72     72      0      0 
LGTC_HiSeq.q                      0.64      1      0     23     24      0      0 
MolEpi_HighMem.q                  0.06      7      0      5     12      0      0 
all.q                             0.59    287      0    157    444      0      0 
mpi.q                             0.61      0      0      0     27      0     27 
notebook.q                        0.00      1      0     23     24      0      0 
qlogin.q                          0.00      3      0     37     40      0      0 

Or

vill@shark:~$ qconf -sql

Prints only the available queues

12hr-KG.q
2hr-all.q
LGTC_HiSeq.q
MolEpi_HighMem.q
all.q
mpi.q
notebook.q
qlogin.q

Currently these queues are available for the following purposes:

queue name	Description
all.q	Everyone can use this queue only the basking shark cannot be used with this queue, can only be used for batch jobs
qlogin.q	Everyone can use this queue, this queue is only for interactive jobs, and runs on koala and kiribati
notebook.q	Everyone can use this queue for their ipython notebook IPythonNotebook

Find the executions hosts for a queue (in this example the LGTC.q) the first command finds the hostlist for the queue.

qconf -sq all.q |grep hostlist | sed 's/[ ]\+/:/g' | cut -d : -f2

This will show : @allhosts

The second command will print all the hosts in the hostlist

qconf -shgrp @allqueue

group_name @allqueue
hostlist baskingshark.researchlumc.nl blueshark.researchlumc.nl \
         camouflageshark.researchlumc.nl caribbeanshark.researchlumc.nl \
         chimerashark.researchlumc.nl cowshark.researchlumc.nl \
         dogfishshark.researchlumc.nl elephantshark.researchlumc.nl \
         epauletteshark.researchlumc.nl frilledshark.researchlumc.nl \
         goblinshark.researchlumc.nl greatwhiteshark.researchlumc.nl \
         greenlandshark.researchlumc.nl kitefinshark.researchlumc.nl \
         lemonshark.researchlumc.nl leopardshark.researchlumc.nl \
         makoshark.researchlumc.nl megamouthshark.researchlumc.nl \
         nightshark.researchlumc.nl pygmeshark.researchlumc.nl \
         reefshark.researchlumc.nl sawshark.researchlumc.nl \
         swellshark.researchlumc.nl threshershark.researchlumc.nl \
         tigershark.researchlumc.nl whaleshark.researchlumc.nl \
         whorltoothshark.researchlumc.nl wobbegongshark.researchlumc.nl \
         zebrashark.researchlumc.nl

Print all hosts in every hostlist

for i in `qconf -shgrpl` ; do qconf -shgrp $i ; done

Which queue do I use?

I you don't give a queue to run on with your job the Open Grid Scheduler will use an available queue where you have access to. If you use the -q queue name option then you use a so called hard_queue list and that queue will be used.