Host connection issues with the VMs.

T

TheChadd

New Member
Oct 8, 2025
2
0
1
Hello,

I have an odd issue with the host connection timing out on with the VMs. If I run an iperf3 test from the host to the VMs, the 1st attempt connects, but the following events respond with 0.00 bites/sec. I 1st noticed the issues when trying to perform a restore and noticed the connection timing out after 20 minutes.

I have run the test with the host to other items on the LAN and WAN, with no issues. Same with iperf3 to the VMs from other devices Pings and traceroutes all look good. I installed an OCP card and tested on it, with the same results. I also tried a different switch, just in case.

It really is just between the host and the VMs on it, which is very strange.

I am on version 9 of PVE, but 8.4 of PBS.
 
Yes, I was and have reset it all back to as default as I can remember. Just eno1 with a single bridge. The firewall was active on all of the VMs, originally, but I have since disabled it.

I did set the MTU to 9000, but that did not help, so it is back to 1500.
 
Hi, please share more information as following if possible:
  1. Host network config: post the relevant parts of /etc/network/interfaces (or whatever file you use) showing eno1, the bridge (vmbr), and how they’re connected.
  2. VM NIC info: inside the VM, what NIC model is shown (e.g. virtio, e1000, etc.)?
  3. Offload settings: run

    Code: 
    ethtool -k <vm_iface>

    inside the VM, and also on the host side (for the tap/bridge interface corresponding to that VM). Paste the full output.
  4. Packet capture: after your first iperf run succeeds, and then the throughput drops, capture traffic on both host and VM (e.g. using tcpdump) to see whether packets are being sent, dropped, or blocked.
  5. Try disabling offloading: on both VM and host interfaces run something like
    Code: 
    ethtool -K <iface> tso off gso off gro off lro off
    and rerun the tests.
  6. Check logs: share any kernel / dmesg / journal entries around the time throughput falls (on both host & VM).
  7. (If possible) test with a different kernel (or fallback kernel) to see if behavior changes.
 
You must log in or register to reply here.
Top Bottom