On 4/3/25 12:34, Markus Frank wrote:
Virtio-fs is a shared file system that enables sharing a directory
between host and guest VMs. It takes advantage of the locality of
virtual machines and the hypervisor to get a higher throughput than
the 9p remote file system protocol.

build-order:
1. cluster
2. guest-common
3. docs
4. qemu-server
5. manager

I did not get virtiofsd to run with run_command without creating
zombie processes after stutdown. So I replaced run_command with exec
for now. Maybe someone can find out why this happens.

changes in v15:
* removed announce-submounts option altogether and always set it for
  virtiofsd.
* added die if both memory hotplug and virtiofs are enabled
* added fstab entry example in the docs
* assert_valid_map_list: only run assert_valid for the mappings on the
  current node
* see individual patches

Gave this another look and spin with the same setup as yesterday (single node and the three mentioned Debian/Fedora/Win11 VMs) and a 3 node test cluster to also test the remaining parts in a cluster setting with migration.

=== Migration tests ===

Tested that offline migration and live migration is blocked correctly, if the resource is not available on another node, i.e. whether there is a dir mapping on another node.

If I try to live migrate a VM from one node to another, where the target does not have a dir mapping defined, I get the following precondition errors in the WebGUI:

- Can't migrate VM with local resources: virtiofs
- Can't migrate running VM with mapped resources: virtiofs0, virtiofs1
- Mapped Resources (virtiofs1) not available on selected target

And on the CLI:

- can't migrate VM which uses local devices: virtiofs

If I try to migrate a VM from one node to another, where the target does not have a dir mapping defined, I get the following precondition errors in the WebGUI:

- Mapped Resources (virtiofs1) not available on selected target

And on the CLI:

- can't migrate to 'node2': missing mapped devices virtiofs1

=== Config tests ===

Otherwise, I did the same tests with xattr / POSIX ACLs as before and this worked just as before. Also the existence of the dir mapping path on the current node is asserted correctly, but in a multi-node setting this could still be improved, see my comments inline.

Otherwise, there were some undefined variables warnings and some warnings that could be improved, I also noted them in my comments inline.


_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel

Reply via email to