Difference between revisions of "Shutting Down WorkerNodes"
From PDP/Grid Wiki
Jump to navigationJump to search| (One intermediate revision by the same user not shown) | |||
| Line 7: | Line 7: | ||
If you prefer to use IPMI, execute | If you prefer to use IPMI, execute | ||
| − | ipmitool -P <PASSWORD> -U root -H host.ipmi.nikhef.nl chassis power off | + | ipmitool -P <PASSWORD> -U root -H ''host''.ipmi.nikhef.nl chassis power off |
| − | where <PASSWORD> is the IPMI management password. The advantage of this solution is that it | + | where <PASSWORD> is the IPMI management password for ''host''. The advantage of this solution is that it switches the node off via its power supply, which will work irrespective of the state of the OS nd its services. However, this does not allow the OS to shutdown properly, so data corruption may occur at the disks. |
Latest revision as of 14:10, 24 April 2009
The simplest method is to use a execute 'shutdown -h now' on all nodes, for example using the following statement from a UI or the PBS server:
for n in $(pbsnodes -a | grep ^wn-) ; do echo $n ; ssh root@$n 'shutdown -h now' ; done
Note that this assumes that the host names of all worker nodes follow the naming convention
wn-<cluster>-<idx>
and that the nodes must have a running ssh daemon.
If you prefer to use IPMI, execute
ipmitool -P <PASSWORD> -U root -H host.ipmi.nikhef.nl chassis power off
where <PASSWORD> is the IPMI management password for host. The advantage of this solution is that it switches the node off via its power supply, which will work irrespective of the state of the OS nd its services. However, this does not allow the OS to shutdown properly, so data corruption may occur at the disks.
