Hi, we’re looking to move our PBS server/scheduler to a new node. What’s the best way to accomplish this with minimal downtime? Would we need to make a failover node and promote that to a primary?
Nodes have to be pointed to the new server (different hostname for the pbs server), hence it is better to pick a specific maintenance day , infrom the users and perform the transition to the new server host.
There are other ways, if you are ready to go this way
- Peer Scheduling
- PBS Pro failover setup
But it would need shared $PBS_HOME and other pre-requisites, it would be like taking 20 steps instead of 5 steps.
Hi
It’s a lot of work to setup a failover. Just schedule the downtime. It’s also hard to test if your failover would really work. When I did a migrate I didn’t set a keepout time and allowed users to run right up to that time. This is better for users that have big jobs that have checkpoints. Consider also if you need to copy across the accounting logs as users might want to query past jobs.
Mike