One scenario that I've been asked to look at is a maintenance scenario. Here's what we're looking at: Say you have server A which is running an active node of a HA pair and server B which is the running backup node of an HA pair. Both servers A and B need an OS patch which will require a complete power cycling of the servers.
Now I realize that if I have set <failover-on-shutdown> to true then I can shut down B, do my OS patch, bring it back up and then shut down A and B will take over. The problem is that what we see in these failover scenarios is that there's about a twenty second pause in both the producers and the consumers during the event. While this is acceptable in a genuine failover event, what my boss would like to see in the maintenance scenario is no interuption of service at all.
Is there anything an administrator can do to prevent this in a maintenance scenario?
in short no, this time is the time taken to start the backup (reading in the journal so journal size will have an effect) plus the time taken for any session state to be recreated.