Cluster node join after reboot troubles

[ott]

Hi,

I have trouble getting node 2 to rejoin cluster after reboot.

I found that neither nodes had the other node defined in "hosts"

GUI>System>Hosts

127.0.0.1 localhost.localdomain localhost
10.0.0.21 pve2.localdomain pve2
10.0.0.135 pve.localdomain pve

So I added pve to pve2 and vice versa (as shown above).

Now they "talk", but I still have to service pve-cluster restart to get all green lights?

 

[mira]

Could you provide the journal of the last boot for both hosts?
journalctl -b > journal.txt and attach the resulting file here.

 

[ott]

Heres the one from pve2. The other one is 17 MB and to large for upload. node pve has not been rebooted.

 

[mira]

Dec 17 16:06:00 pve2 kernel: vmbr0: received packet on ens3 with own address as source address (addr:00:a0:98:71:23:f0, vlan:0)
Dec 17 16:06:00 pve2 kernel: vmbr0: received packet on ens3 with own address as source address (addr:00:a0:98:71:23:f0, vlan:0)
Dec 17 16:06:00 pve2 kernel: vmbr0: received packet on ens3 with own address as source address (addr:00:a0:98:71:23:f0, vlan:0)

Do you have a round-robin bond configured?

It seems time jumped backwards 1 hour. Could it be that time is not synced between the nodes?

 

[ott]

Do you have a round-robin bond configured?

Not intentionally. pve2 is a Proxmox-VM running in TrueNAS Scale (oh yes, the irony...). ens3 is the vNIC assigned to ProxmoxVM.

pve2 is a default/clean install and joined to my original proxmox pve.

It seems time jumped backwards 1 hour. Could it be that time is not synced between the nodes?

That's strange... date returns same time on both nodes, but that's now. when all is 'green'.

btw: pve=v6.4-4, pve2=v7.1-7

 

[mira]

Check the output of ip -details a to see if the bond below vmbr0 is in round-robin mode. Sadly that's the default if it can't set a different one.

 

[ott]

Check the output of ip -details a to see if the bond below vmbr0 is in round-robin mode. Sadly that's the default if it can't set a different one.

Thanks. I am investigating. What exactly am I looking for in the ip -details a output?

 

[mira]

Look for `balance-rr` in the output for the bond attached to vmbr0.