Description of problem: Request for adding deployment option to director to set rx_queue_size and tx_queue_size in /etc/libvirt/qemu.conf The current libvirt patch, proposed by Michal Privoznik https://www.redhat.com/archives/libvir-list/2018-January/msg00608.html , enables setting rx_queue_size and tx_queue_size in /etc/libvirt/qemu.conf, which would enable libvirt to set <driver rx/tx_queue_size=''/> in VM/Instance xml, if rx/tx_queue_size is not explicitly defined in VM/Instance xml. In case rx_queue_size or tx_queue_size is not explicitly set in /etc/libvirt/qemu.conf or qemu-kvm not supporting queue size (for example tx_queue_size in case of older qemu) and <driver rx/tx_queue_size=''/> not set in VM xml, which would be the case for openstack, rx/tx_queue_size would be set to the default value of 256 set by qemu.This bug would resolve setting rx_queue_size and tx_queue_size on per-host basis. Once libvirt patch is part of a release, will open a separate bug for OSP10, to track back-porting of this feature and libvirt to OSP10. upstream Libvirt patch: https://www.redhat.com/archives/libvir-list/2018-January/msg00608.html
https://github.com/openstack/puppet-nova/blob/master/manifests/compute/libvirt/qemu.pp#L23 This puppet class takes care of updating the qemu.conf values.
After a discussion both downstream and upstream, the libvirt patch mentioned in this bug description was reverted.
Hello Christophe Fontaine , yes this bug can be closed.
*** This bug has been marked as a duplicate of bug 1553088 ***