live migration of vms via full mesh network: could not get IP: multiple addresses configured on local node for network

[abma]

I've setup three nodes with direct connections to each other:

https://pve.proxmox.com/wiki/Full_Mesh_Network_for_Ceph_Server#Method_2_.28routed.29

ceph works already fine with it, the migration via the (slow) public network works, too.
i want the live migrations to use the fast connections, too (25GBit). i set in /etc/pve/datacentr.cfg:

migration: type=insecure,network=10.0.15.0/24

but when i try a live migration from node1 -> node2 , i now get the error:

2021-02-20 12:20:44 [node2] could not get IP: multiple addresses configured on local node for network '10.15.15.0/24'
2021-02-20 12:20:44 ERROR: online migrate failure - remote command failed with exit code 255

the all three nodes have valid hosts entries:

10.15.15.50 node1.local node1
10.15.15.51 node2.local node2
10.15.15.52 node3.local node3

any idea whats wrong/missing in my config?

 

[fiona]

Hi,
what is the output of the following on node2?

ip address show to 10.15.15.0/24 up

 

[abma]

ip address show to 10.15.15.0/24 up
4: enp66s0f0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
    inet 10.15.15.51/24 scope global enp66s0f0
       valid_lft forever preferred_lft forever
6: enp67s0f0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
    inet 10.15.15.51/24 scope global enp67s0f0
       valid_lft forever preferred_lft forever

route -n
[...]
10.15.15.50       0.0.0.0         255.255.255.255 UH    0      0        0 enp66s0f0
10.15.15.52       0.0.0.0         255.255.255.255 UH    0      0        0 enp67s0f0
[...]

ssh to the other hosts via the direct connections work... (i.e. "ssh 10.15.15.50")

i guess the ip detection should notice that the ip addresses are the same.

 

[fiona]

Thanks for reporting, this is indeed a bug. I sent a patch.