drbd - diskless node won't join cluster unless started first

Good day,

I am trying to set up a 3-node DRBD9 cluster

• 2 nodes actually have a disk
• the 2 nodes with disk have a direct link via a cross cable and a switch connection with third node
• third node is diskless and acts as a tie breaker

on all three nodes configuration is quite simple:

global_common.conf

global {
  usage-count yes;
}
common {
  net {
    protocol C;
        allow-two-primaries yes;
        transport "tcp";
  }
}

alice-bob.res:

resource alice-bob {
    volume 0 {
       device           minor 0;
       disk             /dev/zd0;
       meta-disk        internal;
    }

    on alice {
       node-id 0;
    }

    on bob {
       node-id 1;
    }

    on charlie {
       node-id 2;
       volume 0 {
           disk         none;
       }
    }

        connection {
        host alice      address 10.0.0.1:7001;
        host bob        address 10.0.0.2:7011;
        }
        connection {
        host alice      address 172.16.0.1:7002;
        host charlie    address 172.16.0.3:7021;
        }
        connection {
        host bob        address 172.16.0.2:7012;
        host charlie    address 172.16.0.3:7021;
        }
}

If DRBD is started on diskless node before the other two everything is fine. If it is started after them, it is shown as standalone.

relevant messages in kern.log on alice:

Apr 18 14:54:04 alice kernel: drbd alice-bob: Declined by peer bob (id: 1), see the kernel log there

on bob:

Apr 18 14:54:04 bob kernel: drbd alice-bob alice: Preparing remote state change 1639412379
Apr 18 14:54:04 bob kernel: drbd alice-bob alice: Aborting remote state change 1639412379

kernel module:

description:    drbd - Distributed Replicated Block Device v9.0.28-1

What am I missing here? Thanks in advance