Shut Down the System¶
Performing a controlled shutdown of an entire StarlingX system may become necessary if, for instance, you need to physically move the underlying hardware.
For information on restarting the cluster, see Start the System.
Note
There are two applicable cases for the shut down:
System with dedicated storage node
System with controller storage
Check in which of the cases the shut down procedure needs to be performed to avoid errors.
Shut Down the System with Dedicated Storage Nodes¶
Prerequisites
On a system that contains storage nodes, a local console or a BMC console connected to storage with monitor (storage-0) is required so that you can issue a shutdown command in the final step of this procedure.
Procedure
Swact to controller-0.
From the Admin > Platform > Host Inventory page, on the Hosts tab, select Edit Host > Swact Host for controller-0.
Lock and shut down each worker node.
From the Admin > Platform > Host Inventory page, on the Hosts tab, select Edit Host > Lock Host.
From the terminal of the worker node, issue a shutdown command.
$ sudo shutdown -hP now
Wait until the node is completely shut down before proceeding to the next step.
Lock and shut down each storage node except for storage node that has monitor (storage-0).
Use the following commands in the active controller terminal to check which storage node has monitor.
$ source /etc/platform/openrc $ system ceph-mon-list
The storage with monitor (Storage-0) is required as part of the Ceph monitor quorum. Do not shut it down until the controllers have been shut down.
Note
This step applies to Ceph-backend systems (systems with dedicated storage nodes) only.
From the Admin > Platform > Host Inventory page, on the Hosts tab, select Edit Host > Lock Host.
From the terminal of the storage node, issue a shutdown command.
$ sudo shutdown -hP now
Wait for several minutes to ensure Ceph has detected and reacted to the missing storage node. You can use ceph -s to verify that the OSDs on the storage node are down.
Lock and shut down controller-1.
From the Admin > Platform > Host Inventory page, on the Hosts tab, select Edit Host > Lock Host.
From the terminal of controller-1, issue a shutdown command.
$ sudo shutdown -hP now
Wait until the node is completely shut down before proceeding to the next step.
Shut down controller-0.
You cannot lock this controller node, as it is the last remaining controller node.
$ sudo shutdown -hP now
Wait until the node is completely shut down before proceeding to the next step.
Shut down the storage node that has monitor (storage-0).
Note
This step applies to Ceph-backend systems (systems with dedicated storage nodes) only.
You must use a local console or a BMC console to issue the shutdown command.
$ sudo shutdown -hP now
Shut Down the System with Controller Storage¶
Procedure
Swact to controller-0.
From the Admin > Platform > Host Inventory page, on the Hosts tab, select Edit Host > Swact Host for controller-0.
Lock and shut down each worker node except for worker that has monitor.
Use the following commands in the active controller terminal to check which worker has monitor.
$ source /etc/platform/openrc $ system ceph-mon-list
The worker with monitor is required as part of the Ceph monitor quorum. Do not shut it down until the controllers have been shut down.
Note
This step applies to Ceph-backend systems (systems with controller storage) only.
From the Admin > Platform > Host Inventory page, on the Hosts tab, select Edit Host > Lock Host.
From the terminal of the worker node, issue a shutdown command.
$ sudo shutdown -hP now
Wait until the node is completely shut down before proceeding to the next step.
Lock and shut down controller-1.
From the Admin > Platform > Host Inventory page, on the Hosts tab, select Edit Host > Lock Host.
From the terminal of controller-1, issue a shutdown command.
$ sudo shutdown -hP now
Wait until the node is completely shut down before proceeding to the next step.
Shut down the worker that has monitor.
Note
This step applies to Ceph-backend systems (systems with controller storage) only.
You must use a local console or a BMC console to issue the shutdown command.
$ sudo shutdown -hP now
Wait until the node is completely shut down before proceeding to the next step.
Shut down controller-0.
You cannot lock this controller node, as it is the last remaining controller node.
$ sudo shutdown -hP now
