Good day and hope you are doing fine.
Currently we are doing NSO HA by issuing command “ha command role-override role slave/master” on both master and slave node to trigger a manual HA. And use command “ ha command role-revert” to switch back.
However, someone recommends to deactivate HA first, then role-override and activate HA again.
May I know what’s the recommended procedure of doing NSO HA?
Thank you so much for your timely response.
However, without issuing this command, in some cases there will be dual active scenario in HA switchover, which is not the expected mode that out customer want to see. May I know the right procedure to do a switchover?
Well the idea is the that a failover occurs when the master fails. You can set the master to not auto start so that the node won't assume its role without the activate command. If a loss of connectivity between master and slave occurs then there could be dual masters. If you controlling the HA mode from an external application you might issue the deactivate after the failover occurs. Some customers use BGP anycast to insure one master always has preference over the other if multiple masters occur.
If you customer requires more complex HA behavior they can also code directly to the NSO HA API as well.
What if I want to do a manual switchover? I mean if there is no failure and just want to trigger a HA switchover? Like a master node replacement?
You can simple deactivate HA on the Master node to initiate the failover.
user@ncs> request ha commands deactivate
Failover will take (interval*failure limit) seconds before the failover slave confirms loss of master and initiates a failover.