When doing things like VM transfers between hosts in a cluster, PVE uses ssh (https://pve.proxmox.com/wiki/Cluster_Manager#_role_of_ssh_in_proxmox_ve_clusters). In my setup, my cluster members have multiple network interfaces, an "admin" interface, and a "data" interface. The admin interface is what matches the host's hostname, and what DNS resolves to, but it's only a 1Gb link. The data interface is a 20Gb link, and what should be used for most activity, including VM transfers between cluster members.
The problem is that PVE is doing these transfers over the admin interface, while I want it to use the data interface. Originally I was thinking I could add an entry in
While I could technically still do the
I've tried swapping the ring0 and ring1 addresses in corosync, but that had no effect.
The problem is that PVE is doing these transfers over the admin interface, while I want it to use the data interface. Originally I was thinking I could add an entry in
/etc/ssh/ssh_config
to override the address used when connecting by name. However when I looked at the ssh being invoked, it's connecting by IP address:
Code:
/usr/bin/ssh -e none -o BatchMode=yes -o HostKeyAlias=ded4461 root@1.2.3.4 -o ExitOnForwardFailure=yes -L /run/qemu-server/100.migrate:/run/qemu-server/100.migrate -L /run/qemu-server/100.migrate:/run/qemu-server/100.migrate /usr/sbin/qm mtunnel
While I could technically still do the
ssh_config
override on IP address, that just smells like a really bad idea. I would prefer if I could either get PVE to use hostnames for ssh, or change the address it uses. However I'm unsure how to accomplish either of these.I've tried swapping the ring0 and ring1 addresses in corosync, but that had no effect.
Last edited: