Force migration via another network

[itvietnam]

Hi,

We have dual ring:

• ring0: 192.168.0.0/24
• ring1: 192.168.1.0/24
• migration network: 192.168.3.0/24

We have defined in /etc/pve/datacenter.cfg with following setting:

migration: secure,network=192.168.3.0/24

Now we do migrate ring0 network of whole cluster (rolling update) from bond1 to vmbr0 (which run on bond0). We added a server with new setup. Then we join cluster via ring1. This server is already hosted ring0 on bond0.

Now we migrate VM from existing node to this new server (called node120) and the log always migrate use ring0 to migrate.

task started by HA resource agent
ssh: connect to host 192.168.0.170 port 22: No route to host

TASK ERROR: command '/usr/bin/ssh -e none -o 'BatchMode=yes' -o 'HostKeyAlias=node120' root@192.168.0.170 pvecm mtunnel -migration_network 192.168.3.0/24 -get_migration_ip' failed: exit code 255

Can we fix above error?

As my thought, we can do live migration network ring0 without have to shutdown whole cluster because ring1 will work as backup. Correct huh?

 

[fabian]

the migration network is only used for the (possibly heavy) migration traffic itself. we need to contact the other node to find out its address in that migration network, and for this you need access via the regular IP.