[SOLVED] Ipv6 issue in HA-Cluster

[dRei]

Hey,

I'm facing some problems in my setup. I've setted up two PM-Gateways and want to combine these as a cluster. Both of the Containers receive an IPv4-LAN-Address and some IPv6-Addresses (ULA, Public, Localhost). When I create a Cluster on the first container I get the IPv4-Address displayed as the IP-Address I have to give into the Mask on the second container but I could not reach the First node from the second node when I enter the IPv4-Address in the Cluster-Config. I could ping the IPv4 as well as the IPv6 addresses in both directions and there is no additional Firewall in this network. So every communication within this LAN should be good. If I try to use ssh to bring the second container in the cluster I get an timeout if I use the IPv4-Address. If I use the IPv6-ULA-Address instead the connection is made and the second container joins the cluster. But the communication between both nodes seemed to be broken in this scenario because in the Cluster Config on the first container there is showing up the ipv4-addresses as endpoints which will not communicate to each other.

Could I change the configuration of the HAManager/Clustermanager to use IPv4 as binding-endpoint or how could I bring this to work?

Thanks, Michael

 

[dRei]

Really nobody? I see this as a bug in PMG because I could not find any solution to bring my PMG-Cluster to live with an dualstack network... I use PMGs with a fresh installation on an nearly fresh installed PVE-Cluster.

 

[Stoiko Ivanov]

last time I tried dual-stacked PMG clusters worked - things I'd check:
* where is pmgproxy and sshd listening on: `ss -tlnp`
* can you ssh between the nodes without any issues on both address families
* any relevant logs in the journal when you try to join the 2 nodes

 

[dRei]

Damn.... followed your tipps and tried to connect via ssh from slave to master fails on ipv4 as well. So I checked the settings and configured my MTU in the container to match the proxmox VE mtu and now it works like expected...