VM/network issues.

On 2/24/2022 at 4:10 AM, Hanfufu said:

I also saw this, under IPv6 there are lots of discarded packets.

I also disabled IPv6 on the VM a while back, to make sure that it didnt cause problems.

Also i posted a screenshot of Network Settings, showing something odd with IPv6 route, but im not a network expert so dont know if its relevant.

 

For whatever reason mentioning ipv6 gave me an odd idea about my VM having an error in the network config. I sort of solved my problem on accident, but I'll still share how I did it for anyone who is trying to solve this problem for linux distros. I am by no means a professional so bare with me... lol

I run headless VMs that only have black terminals so I don't have fancy GUI displays like yall do, but I will try my best to explain the tools I used to prove my issue is related to packet loss and IPV6

I run ubuntu server 22.04

First off I knew I was already having an issue with high packet loss because my ssh sessions would consistently break every 5 minutes or so.

To prove my problem in greater detail I ran an iperf test. On my unraid server I run an Iperf server in a docker. iperf is for benchmarking network bandwidth and transfer rate stuff. I used iperf on the VM to run a bidirectional data transfer test between the VM host and VM guest for 30 seconds.

The test output shows the packets doing as they should until suddenly the connection is dropped by the unraid server for no apparent reason which usually means there is a timeout error going on somewhere. Anytime I observe utterly dog water network performance I quickly consider the probable possibility there is something very wrong with the network configuration on the VM operating system. I ran the test again just to eliminate any false positives. This test was dropped almost immediately so I moved on to investigating the cause for so many packet drops.

There is about a baker's dozen posts about people correlating their VM connection issues with a 100% spike on a single thread in unraid. I believe this correlation is absolutely correct, but in my perspective I see most people get it wrong by treating the spike as a cause of their problems rather than a symptom from an underlying problem. To me it makes perfect sense that a fully saturated cpu thread would happen due to a misbehaving virtual network adapter. I strongly think unraid is forcing the VM network adapter to crash repeatedly because the over saturated thread simply aborts all pending tasks when it reaches 100% usage.

I'm not really sure why my mind jumped to this, but I suddenly remembered reading somewhere that virtual network adapters can't handle IPV6. I honestly have no idea why I thought of this, but I took a look-see at the Netplan configurating for ubuntu. In there I discovered I mistakenly enabled IPV6 DHCP on the VM. I commented out the line for DHCP IPV6 and applied the configuration. 

After restarting the NetworkManager my packet loss issue was completely gone. It seems in my case the VM was constantly making DHCP requests to the network gateway, but due to some issue with IPV6 addresses the virtual network adapter blocks the VM from acquiring the address it asked for. Essentially this puts the VM and unraid into a infinite loop of DHCP requests that the virtual adapter will block, over-saturating the cpu thread, crashing the virtual network adapter, then dropping all active connections to and from the VM. The loop starts again when the virtual network adapter restarts, allowing the VM to make another DHCP request for an IPV6 address to the gateway.

That is my best explanation, but its probably not 100% accurate. Hopefully some of yall will find this information useful.

Edited July 22, 20241 yr by JD1099