Sometimes it can be desirable or necessary to set the corosync timeout for the PCS cluster higher.
This can, for example, be needed in environments where backups or snapshots of the virtual machine(s) regularly break the cluster due to the VMs being halted / frozen whilst the VM is "quiesced" (brought to a consistent snapshottable state).
The default value is 1000ms.
You can verify the current timeout like this:
corosync-cmapctl | grep totem.token |
One one of the load balancers, edit the /etc/corosync/corosync.conf file.
Add the following entry to the totem section.
token: 4000 |
The 4000 here will increase the token timeout to 4000 ms.
THIS WILL NOT WORK IF A PROXY IS BEING USED IN THE NETWORK. You will have to edit the configuration file on all nodes instead. |
pcs cluster sync |
pcs cluster reload corosync |
On all nodes:
corosync-cmapctl | grep totem.token |