[SOLVED] Error 596 : Issue with communication between two nodes (1 and 2), but no issues between nodes 3 and 4

[sbe]

Hello everyone,

I'm experiencing a communication issue between two of my Proxmox nodes (1 and 2), while everything works fine between nodes 3 and 4.
The communication test is being performed through the web GUI using the SHELL menu.

Here’s a summary of the situation:

• pve 1 => 2: ERROR 596
• pve 1 => 3 and 4: OK
• pve 2 => 1: ERROR 596
• pve 2 => 3 and 4: OK
• pve 3 => 1, 2, 4: OK
• pve 4 => 1, 2, 3: OK

In the shell, I’m getting the following errors:

kex_exchange_identification: read: Connection reset by peer
Connection reset by <pve-ip> port 22

I’m also seeing this error in the GUI:
Connection error 596: Connection reset by peer

Already tested:
I’ve tried connecting via SSH using the following command:

ssh -o "HostKeyAlias=pveX" root@<pve-ip>

Note: An upgrade was recently done on pve 1 and pve 2 via the GUI.

pve1 = v8.2.7
pve2 = v8.1.4
pve3 = v8.1.4
pve4 = v8.1.4

Any advice or troubleshooting steps would be greatly appreciated.

Thanks in advance!

 

[Pifouney]

maybe a time synchronisation problem

 

[sbe]

Thank you so much for your suggestion, I really appreciate your help!

I performed a manual test to check time synchronization across the servers:

root@pve1:~# date
Thu Oct 10 02:30:28 PM CEST 2024

root@pve2:~# date
Thu Oct 10 02:30:28 PM CEST 2024

root@pve3:~# date
Thu Oct 10 02:30:29 PM CEST 2024

root@pve4:~# date
Thu Oct 10 02:30:29 PM CEST 2024

After 24 hours, the issue has evolved without any maintenance operations on the nodes.

• pve2 => pve1: Now working fine (there was a request to update the SSH certificate).
• pve1 => pve2: I’m still getting an error in the shell interface: Host key verification failed.
  However:
• pve1 => pve2 works fine for all other menus.

After running the command ssh -o 'HostKeyAlias=pve2' root@X.X.X.12 from pve1, everything is working again.

Could it be that after running several commands, we just needed to wait for a few hours? What’s your opinion on this?
Does that look good to you?

I'm changing the status of this issue, even though the root cause wasn’t identified.

Thanks in advance!

 

[Pifouney]

I d'ont have any other idea, maybe found in logs why before they failed to communicate :/

Happy to see that 's solved