TOPIC:

The number of hosts in your pool is being detected as 1. Probably because you are putting your master into maintenance mode or performing a graceful shutdown of the master to simulate your test - neither of which is a valid DR scenario for testing.

Run your test again making sure that you have a complete operational pool and then crash your master.

Thank you, yes that seems to be the case.

How would I recover from the situation like that?
( get the slave which is still running to become master and take over ?)

btw: if we detect that only one server is in the pool should not that one server be promoted to master and take over?

Generally that is not a situation to recover from. If the master is to be gracefully shutdown, then HA should be disabled before the shutdown.

However, forgetting to disable HA before a graceful shutdown should not cause the slave to self fence. Ideally it should do nothing in order to protect the user and any running VMs. We'll make a code change so that this scenario is better dealt with.

Thank you.

Short term what would be the procedure to recover from it?

We are considering a small change that would prevent a slave from fencing while its peer is in maintenance mode. This will NOT allow a slave to become a master while the master is in maintenance mode. The solution for your scenario is to disable HA if you are going to put the master into maintenance mode.

Thank you.
Lets assume that I made a mistake and the master went down gracefully without me disabling the HA and I am in this situation. How do I recover?