virtio-net-tx-queue-size reflects in nova conf but not for the vm even after a hard reboot

[Nishant Dash]

After modifying the nova compute config options,
- virtio-net-rx-queue-size=512
- virtio-net-tx-queue-size=512

I hard rebooted my vm and spawned a new vm and what I see (on both of them) is:
- virsh xml
```
# virsh dumpxml 2 | grep -i queue
      <driver name='vhost' rx_queue_size='512'/>
```

- nova.conf
```
# grep -i queue /etc/nova/nova.conf
tx_queue_size = 512
rx_queue_size = 512
```

- inside the vm
```
root@jammy-135110:~# ethtool -g ens2
Ring parameters for ens2:
Pre-set maximums:
RX: 512
RX Mini: n/a
RX Jumbo: n/a
TX: 256
Current hardware settings:
RX: 512
RX Mini: n/a
RX Jumbo: n/a
TX: 256
```

The RX config gets propagated, but the TX config does not
Please let me know if any more information is needed.

----------------------------------------------------------

env:
- focal ussuri
- nova-compute:
    charm: nova-compute
    channel: ussuri/stable
    revision: 669
- this is a freshly deployed openstack on vms (not on baremetal)
- libvirt: 6.0.0-0ubuntu8.16
- nova-compute-libvirt 21.2.4-0ubuntu2.5
- qemu 4.2-3ubuntu6.27

[Billy Olsen (billy-olsen)]

This does not appear to be a charm issue, but rather it appears to potentially be a nova issue. I can confirm that setting the rx_queue_size and tx_queue_size results in the nova.conf file being updated by the charm, but that the resulting hard rebooted guest does not get the tx_queue_size, only the rx_queue_size.

[Rafael Lopez (rafael.lopez)]

Based on libvirt docs [1], the device type (not driver) may not be supported. TX only seesm to be supported by the 'vhostuser' type device:

"tx_queue_size
The optional tx_queue_size attribute controls the size of virtio ring ..... For instance, QEMU v2.9 requires value to be a power of two from [256, 1024] range. In addition to that, this may work only for a subset of interface types, e.g. aforementioned QEMU enables this option only for vhostuser type. "

Nova actually only sets tx when the device is type 'vhostuser', via the _set_config_VIFHostDevice() codepath:
https://opendev.org/openstack/nova/src/commit/f0565e84ee9578d6dafd22d57fb0c95cb3984c1e/nova/virt/libvirt/vif.py#L533

[1] https://libvirt.org/formatdomain.html#setting-nic-driver-specific-options

[Felipe Reyes (freyes)]

Marking charm-nova-compute task as invalid since it is not at fault here (neither nova) based on Rafael's comment. If you disagree, please reopen and state what we needs to be fixed.

[Launchpad Janitor (janitor)]

[Expired for OpenStack Compute (nova) because there has been no activity for 60 days.]