@kassebasse said in Remove running host:

I am just trying to remove the host, because it is no longer exists

In that case, try

xe host-declare-dead

You will need the prior UUID of the host, that is, the UUID assigned to the host before you reinstalled xcp-ng.
xe host-list will provide that.

Read the warnings about host-declare-dead carefully.

The existing technical documentation is great. An operations guide would be helpful. Here are a couple of chapter ideas:

1. Best practices for setting up your environment
   (eg, odd number of hosts, isolate management network, treat your hosts like cattle, etc)

2. Preparing for disasters (eg, requirements for restoring pool metadata, how to recover a single VM if your pool data is gone, etc)

@olivierlambert
Can we rule out extra_guest_irqs as the root cause of this problem?

https://docs.xcp-ng.org/compute/#nvme-storage-devices-on-linux

@olivierlambert
Is the CPU on the sending host or the receiving host the limiting factor for single disk migrations?

Je le connecte à une machine virtuelle Debian 12, noyau 6.1.0-31-amd64. Le disque est un Western Digital "MyBook" que j'utilise pour les sauvegardes hors ligne.

@kagbasi-ngc @olivierlambert
I've seen similar problems with other LDAP clients. With those other systems, it was a problem with a fixed buffer size somewhere in the client code. If the list of groups exceeds what will fit in the buffer, we had weird problems. Increasing the buffer size fixed the problem.

C'est intéressant. J'ai un disque USB de 14 To connecté directement à l'hyperviseur. Il est attribué à une VM via VUSB, et il fonctionne correctement sur la version 8.2.1 de XCP-ng.

@cairoti
You might look at "Appliances".
The startup delay on the VM doesn't apply when the host is booting, only when you run "xe vm-start".

Look at
https://xcp-ng.org/forum/topic/6464/resolved-vm-s-auto-starting-in-wrong-order
and
https://xcp-ng.org/forum/topic/1312/vm-start-delay-does-it-work-yet/5

The existing technical documentation is great. An operations guide would be helpful. Here are a couple of chapter ideas:

1. Best practices for setting up your environment
   (eg, odd number of hosts, isolate management network, treat your hosts like cattle, etc)

2. Preparing for disasters (eg, requirements for restoring pool metadata, how to recover a single VM if your pool data is gone, etc)

@kassebasse said in Remove running host:

I am just trying to remove the host, because it is no longer exists

In that case, try

xe host-declare-dead

You will need the prior UUID of the host, that is, the UUID assigned to the host before you reinstalled xcp-ng.
xe host-list will provide that.

Read the warnings about host-declare-dead carefully.

@kassebasse said in Remove running host:

Hi!
I've gotten myself into a bit of a mess I guess and I need help to resolve it. I changed storage on one of my host, and that meant that I did not power down the host correctly. But now, when I try to forget the host, a running VM does not let me remove the host.

The problem:
The host is "running" even though it is not connected.
It has a "running" ghost VM
The VM has a GPU attached
Trying to shut the VM down results in the host offline error message.

How can I resolve this issue?

@kassebasse , please clarify a few things:

1. When you sat that the host is "running", do you mean that the host is powered on?
2. Is there an SR on the host?
3. Is there any data on the host that you want to keep?
4. What are you trying to accomplish by forgetting the host?

On Fedora 39, you must enable the daemon manually after you install the package. This is probably true on all RPM-based systems.

$ sudo systemctl enable --now xen-guest-agent.service

I'd be happy to submit a pull request to add that to the RPM package if you will point me at the right repo.