Issue with Windows VM on XCP-NG

Computer Hardware & Server Infrastructure Builds
1

Been using XCP-NG on our servers here at work for a few years now after moving off of hyper-v. We run exchange in house here, and in running the health checker script I noticed an error about the packets being discarded. I tried both the xcp and the citrix drivers but the result is the same. There are no NIC power management settings in any of the tabs and I did run the powershell command to disable nic power management just in case but the errors persist. I booted up one of the VMs on the old hyper V server as a test and it did not have these errors. The event lots also don’t have any useful info.

I also tested the MTU and it appears 1500 is correct (1472+28).

Screenshot: https://imgur.com/7SlWFt3

I haven’t noticed any problems in production but having all the checks in the green would be nice. Any thoughts would be appreciated!

2

What health checker script are you running?

3

It’s a script from Microsoft that gives you a general overall status of your exchange servers.

https://microsoft.github.io/CSS-Exchange/Diagnostics/HealthChecker/

This is the check that it is failing on:

https://microsoft.github.io/CSS-Exchange/Diagnostics/HealthChecker/PacketsLossCheck/

We haven’t had any issues, and I was able to replicate the issue on another server (Both Dells, a T430 and an r730).

Last night I turned off offloading using this script, we’ll see what happens:

4

Not something I have tested for or had an issue with. Maybe someone here will have some more insight or you could post inthe XCP-NG forums.

5

Yeah, I’ll try posting there. And will update this thread if something useful comes around.

6

I’d try this first:
netsh int tcp set global autotuninglevel=disabled

saves my ass frequently
turn all back:
netsh int tcp set global autotuninglevel=normal

7

Just tried it and sadly still getting the packet loss in the performance counter:

I posted in the XCP-NG forums as well, will post back here if there are any updates.

image
image1004×382 31.7 KB

8

Just to follow up on this, I have tested with some other servers and have found it is only the BROADCOM network adapters having this issue. Crappy Broadcom NICs in dell servers strike again.

1 Like
9

Actually aligning MTU size obviously the main reason of many miracles. Just fought with bunch of it.
Than we are in bridge/vlan space it may decreases a little bit.