| Scyld ClusterWare HPC: Reference Guide | ||
|---|---|---|
| << Previous | Next >> | |
This script can be used to bring a node down ("reboot", "halt", "pwroff") in such a way that the local filesystems on the compute node remain in a constant state.
node_down works by first changing the node's state to "unavailable", then remounting all of the filesystems read-only, followed by using bpctl to perform the actual state change you requested.
The node number of the node to bring down.
The state to put the node in after remounting all the filesystems. The state defaults to "reboot" if unspecified.
Cleanly rebooting node 3:
[root@cluster ~] # /usr/lib/beoboot/bin/node_down 3 Remounting / readonly... Remounting /proc readonly... Remounting /home readonly... Remounting /dev/pts readonly... Synching disks. Shutting down node 3 (reboot) |
| << Previous | Home | Next >> |
| bpslave | Up | recvstats |