We’ve setup a three node cluster with DRBD 9.0.12 where all three nodes are primary and UpToDate.
This setup is working fine until we reboot one machine. After the machine comes up again, it won’t reconnect. It’s state remain in „Outdated“ while it tries to connect to the other hosts (which are still both primary).
It does not matter what drbdadm command we execute. The state won’t change. drbdadm up , drbdadm connect or drbdadm --discard-my-data connect
The only thing which works as workaround is putting one of the other two primarys to secondary and then disconnect and reconnect them. After this the rebooted host will connect and start syncing. But in a real world scenario this is not practicable to down a resource on one of the both survivors.
What's the right way after node failure in a tripple primary setup?
Our second try was to configure a dual primary plus one secondary setup. But the problem is still the same.
Is it supported and/or possible to have a primary/primary/primary or a primary/primary/secondary setup with DRBD 9.0.12? The documentation is not very clear at this point