Before you proceed with the procedure, verify that all backups are up to date and, if possible, are stored externally.
If you have Ceph in your MCP cluster, power it off as described in Shut down a Ceph cluster.
To shut down an MCP OpenStack environment:
Power off the prx
nodes to discontinue all external connections:
salt 'prx*' system.poweroff
Power off the OpenStack bare metal bmt
controller nodes:
salt 'bmt*' system.poweroff
Power off the ctl
nodes:
salt 'ctl*' system.poweroff
If OpenContrail is deployed, power off the ntw
and nal
nodes:
salt 'ntw*' system.poweroff
salt 'nal*' system.poweroff
Power off the gtw
nodes:
salt 'gtw*' system.poweroff
Power off the msg
nodes:
salt 'msg*' system.poweroff
Power off the dbs
nodes in the following strict order:
salt 'dbs03*' system.poweroff
salt 'dbs02*' system.poweroff
salt 'dbs01*' system.poweroff
Power off all StackLight LMA nodes:
salt 'log*' system.poweroff
salt 'mon*' system.poweroff
salt 'mtr*' system.poweroff
If you have any containers running on the cid
nodes,
for example, Jenkins, stop them.
Power off the cid
nodes:
salt 'cid*' system.poweroff
If local Aptly is used, power off the Aptly node.
salt 'apt01*' system.poweroff
Power off the cmp
nodes:
salt 'cmp*' system.poweroff
Verify that all control plane instances are down except for the
cfg01
node if the latter runs on a KVM node:
salt 'kvm*' cmd.run 'virsh list | grep -v cfg01'
Power off the kvm
nodes.
Since the cfg01
node usually runs on a KVM node, postpone the KVM nodes
shutdown, for example, for 15 minutes, to safely shut down the cfg01
node:
Warning
After running the command below, you will lose any access to the environment in 15 minutes.
salt 'kvm*' cmd.run 'shutdown --poweroff 15' --async
Power off the cfg01
node before the kvm
nodes have powered off:
salt 'cfg01*' system.poweroff