[DRBD-user] Problem with a three-node configuration

[Robert Altnoeder]

On 3/18/19 5:44 PM, JCA wrote:
> /etc/drbd.d/some-data.res:1: in resource knot-data:
> There are multiple host sections for the peer node.
> Use the --peer option to select which peer section to use.
> resource some-data: cannot configure network without knowing my peer.
>
> I had a look into the drbdadm man page, and it says that, when more
> than two host sections are in play, one must use the --peer option to
> specify the peer node to connect to. So I tried the following:
>
> # drbdadm up --peer two some-data
> Device '1' is configured!
> Command 'drbdmeta 1 v08 /dev/sdb1 internal apply-al' terminated with
> exit code 20
>
> Any ideas on what is going on, and how to fix it?

You are trying to run a 3-node configuration, which is only supported
with DRBD 9.x, with DRBD 8.x, and you are also trying to run a dual
primary setup in a 3-node configuration, which works only temporarily
(e.g. for live migration of virtual machines on a hypervisor), but is
otherwise only fully supported in a 2-node setup.

Upgrade the kernel module to DRBD 9.x, correct the configuration file
(DRBD 9.x requires a node-id in the "on" section), and turn off dual
primary mode unless you are absolutely sure you know exactly what you
are doing and you have a reliable fencing mechanism.

br,
Robert