ShutDownCluster
How to shut down the cluster
Kill all remaining jobs on the cluster and close the queues
Go to cream02 and issue
./kill_all_jobs
Wait for all jobs to terminate. This will ensure that when one issues the shutdown of the worker nodes, there will be no lock on the /scratch
To close the queues:
cd ./disable_queues
Switch off the worker nodes
on ccq:
cd distrib-scripts/ ./distrib_exec_list node_list_all_wns "poweroff"
verify that all the WN's are indeed in the file node_list_all_wns. This file should be kept up-to-date.
Switch off storage
still on ccq
./distrib_exec_list list_all_storage "poweroff"
Also verify the file list_all_storage for completeness.
Next, you still need to switch off the disks of some of the storage machines. You need to be at the CC to do this.
Switch off some servers
This will shut down a few virtual servers that need a proper shutdown
./distrib_exec_list list_servers_proper_shutdown "poweroff"
Switch off the rest of the machines
mainly the M machines
./distrib_exec_list list_proper_shutdown_the_rest "poweroff"
Last switch off
power off ccq
verify
verify that all is indeed off