Proxmox loosing connectivity when VM starts network activity

[damjank]

Hello friends!

I have PVE 6.4 on HP DL560 G8. I have integrated NIC in use (quad giga port). Now my problem is this - when I installed PVE, it was all fine, I could get IP via DHPC, which I then reconfigured to be fixed. I could update system etc. etc. - so no problems accessing PVE or working with it.
Then I created 1 sample VM, using Debian 10.7 ISO. The virtual HW for it was as 4x vCPU, 16GB RAM, 40GB DISK (disks in PVE are Samsung Ent SSDs).
I started VM, which came up normally, started the installer and as soon as I got to the screen with network config, I selected DHCP, naturally, and immediately after that, I got disconnected from PVE, and as I would imagine also the VM gets disconnected.

After some 20 minutes, I can then connect back to the PVE, and try to resume with VM but any additional network activity from it - same stuff - network gets disconnected.

Now for everyone - the network is connected to Cisco Nexus 3900, and there are already other hypervisors, running several dozen VMs without any trouble, so I am almost ruling it out (for now) doing some port disable or similar - but I am open for suggestions if that may not be the case.

Apart from that, any pointes, hints, suggestions?
Diagnostics data is in attachment of this thread.


Thank you in advance.
Good day!
D

 

[6uellerbpanda]

maybe duplicate ip or mac addr?

what is syslog on pve node and switch telling when it happens?

can you post your interface config from the switch?

 

[damjank]

I checked, no duplicates were found - there are 4 NET adapters, only 1 is participating in virtual switch, the switch has mac of that adapter, VM has its own, other VM on net have completely different MAC starting addresses.
PVE does no information in syslog at all, when disconnect happens, all there is is this:

May 05 11:04:00 pve1 systemd[1]: Starting Proxmox VE replication runner...
May 05 11:04:01 pve1 systemd[1]: pvesr.service: Succeeded.
May 05 11:04:01 pve1 systemd[1]: Started Proxmox VE replication runner.
May 05 11:05:00 pve1 systemd[1]: Starting Proxmox VE replication runner...
May 05 11:05:01 pve1 systemd[1]: pvesr.service: Succeeded.
May 05 11:05:01 pve1 systemd[1]: Started Proxmox VE replication runner.
May 05 11:06:00 pve1 systemd[1]: Starting Proxmox VE replication runner...
May 05 11:06:01 pve1 systemd[1]: pvesr.service: Succeeded.

I am trying to get someone to get me the config as I am not at the office.
To recap - PVE does no information. I will setup some kind of network check so that I at least see if it get disconnected or am I that gets disconnected :S strange behaviour indeed - since there are more hypervisors on SAME network equipment...

 

[6uellerbpanda]

I checked, no duplicates were found - there are 4 NET adapters, only 1 is participating in virtual switch, the switch has mac of that adapter, VM has its own, other VM on net have completely different MAC starting addresses.
PVE does no information in syslog at all, when disconnect happens, all there is is this:

when you say virtual switch it means vmware isn't it ? does that mean you're running proxmox on a vmware host ?

 

[damjank]

No, it means, that on this physical Cisco switch, there are more than 1 hypervisor connected. They all run VMs and all are working properly, no disconnects or similar problems. On this very same device, I added PVE6.3, fired it up, connects OK, but immediately when first VM starts *ANY* network operation, eg. DHCP, NTP, similar, I loose connection to PVE for about 20min - but all other hypervisors and their VMs are OK durring that outage... so there is definitely something going on the PVE side