[SOLVED] Proxmox 8 | Sporadic ICMP / Scheduler Problems

[fstrankowski]

I've setup a fresh Proxmox 8 Box at Hetzner DC and have a new kind of problem with it i never had before. After setting up the box, i used to outward ping hosts to check the connectivity from time to time and its link latency - which is fine. But while pinging servers, waiting for the reply made me stumble upon the fact that _some random_ icmp-replies are listed slower than others.

https://share.nordee.de/nordee/23/07/_pn_HEa3Y6JUBmiLmXPVDE7zN3WUycdaEaVk.mp4

If you take a look at this video, after 00:50 seconds (icmp seq 50+) you can see that some replies are popping up at the same time while having no abnormal latency. So it looks like a scheduling setting prevented the "in time display" of the reply for any reason. This only happens for pings i do on the host itself, pinging the host from an external source does _not_ show this behavior.

Also testing this with a clean Debian Buster/Bullseye/Bookworm did not show this behavior, neither did the Hetzners rescue image. Here are some facts:

• Minimal network config (nothing special, same routes as Hetzners rescue image)
• No Firewall enabled
• No VMs enabled
• No modifications to Proxmox. Clean fresh Proxmox 8 ZFS Raid 1 install
• NVMe harddrives are in good shape and working
• No load on the host itself

Did Proxmox change anything scheduler related? This phenomena is absolutely random and i can only replicate it by testing it over and over.

Florian

 

[fstrankowski]

I've found the problem: Reason for the behavior has been a faulty DNS server in the line. This caused the resolving reverse names for the ipv6 adresses to be slower than 1 second and thus caused the problem shown above.