Locked History Actions

OPS/FAQs/Torque_Maui_Reservations

Introduction

  • We outline the necessary steps taken to configure the reservation of 2 slots under queue grid.q. The queue is shared between 3 groups of users (ops, opssgm and opsplt) but the reservation will allow immediate job execution from users belonging to opssgm and opsplt groups.

  • We assume that you have configured the previous users and groups otherwise the following guidelines will not work.
  • The implementation of this framework implies two steps:
    1. Changes at the Torque Server level to deploy and configure the queue.
    2. Changes at the Maui Server level to set the resource reservation.

Torque

  • The following shell commands create, enable and configure the grid.q queue with specific cpu and wallclock times, and a max limit on the number of running jobs:

qmgr -c "create queue grid.q queue_type=execution" 
qmgr -c "set queue grid.q resources_max.cput=12:00:00" 
qmgr -c "set queue grid.q resources_max.walltime=24:00:00" 
qmgr -c "set queue grid.q enabled=true" 
qmgr -c "set queue grid.q started=true"
qmgr -c "set queue grid.q acl_groups=opssgm"
qmgr -c "set queue grid.q acl_groups=opsplt"
qmgr -c "set queue grid.q acl_groups=ops"

MAUI

  • The following commands allow the reservation of 2 extra slots allowed to move from host to host over time in an attempt to optimize resource utilization:

echo "SRCFG[sam] PERIOD=INFINITY" >> /var/spool/maui/maui.cfg
echo "SRCFG[sam] TASKCOUNT=2" >> /var/spool/maui/maui.cfg
echo "SRCFG[sam] RESOURCES=PROCS:1" >> /var/spool/maui/maui.cfg
echo "SRCFG[sam] GROUPLIST=opssgm,opsplt" >> /var/spool/maui/maui.cfg
echo "SRCFG[sam] ACCESS=DEDICATED >> /var/spool/maui/maui.cfg
echo "SRCFG[sam] FLAGS=SPACEFLEX" >> /var/spool/maui/maui.cfg
echo "GROUPCFG[opssgm] PRIORITY=1000 >> /var/spool/maui/maui.cfg
echo "GROUPCFG[opsplt] PRIORITY=1000 >> /var/spool/maui/maui.cfg
  • Since the reservations have to be asked by the jobs, we have to force all users from opssgm and opsplt to request the reservation

echo "GROUPCFG[opssgm] FLAGS=ADVRES:sam.0.0" >> /var/spool/maui/maui.cfg
echo "GROUPCFG[opsplt] FLAGS=ADVRES:sam.0.0" >> /var/spool/maui/maui.cfg
  • The maui service has to be restarted to reflect the configuration

/etc/init.d/maui restart

References

  1. Torque Resource Manager Manual

  2. Torque Resource Manager QMGR Manual

  3. Torque Resource Manager Queue Configuration

  4. Maui documentation

  5. Maui managing reservations documentation