as reported in this downstream bug https://bugzilla.redhat.com/show_bug.cgi?id=2040554
os vif currently set the requested mtu as part of a separate request to the port add port command.
as a result the port is initially created with an mbufer pool for the default 1500 size then updated to use the mbuf pool for the jumboframe size.
this results in an error in the openvswitch-vswichd log
Dec 19 15:16:28 cpdh002a-lab3a-vxf ovs-vswitchd[238324]: ovs|2438761|netdev_dpdk|ERR|Failed to create memory pool for netdev vhu3a1c44e6-d7, with MTU 1500 on socket 0: Invalid argument
Dec 19 15:16:28 cpdh002a-lab3a-vxf ovs-vswitchd[238324]: ovs|2438762|dpif_netdev|ERR|Failed to set interface vhu3a1c44e6-d7 new configuration
Dec 19 15:16:28 cpdh002a-lab3a-vxf ovs-vsctl: ovs|00001|vsctl|INFO|Called as ovs-vsctl --timeout=120 -- set interface vhu3a1c44e6-d7 mtu_request=9000
while this error its harmless since the follow call to set the mtu will correct the issue momentary it can be annoying to operators as it create noise in the logs.
the fix for this is relivitly trivial
we just need to move the mtu update into the transaction that creates the port
https://github.com/openstack/os-vif/blob/master/vif_plug_ovs/ovsdb/ovsdb_lib.py#L179-L185
which will require a minor refactoring of
https://github.com/openstack/os-vif/blob/master/vif_plug_ovs/ovsdb/ovsdb_lib.py#L44-L68
to support being called in a transaction.
This bug has no other negative sideffect that we know of at this time and can simply be worked around by ensuring ovs-dpdk has enough memory for both the default mem pool and jumbo frame mem pool which is already considered a best practice
https://developers.redhat.com/blog/2018/03/16/ovs-dpdk-hugepage-memory
https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/13/html-single/network_functions_virtualization_planning_and_configuration_guide/index#c_ovsdpdk-memory-params
Fix proposed to branch: master
Review: https:/