Uploaded image for project: 'Red Hat OpenStack Services on OpenShift'
  1. Red Hat OpenStack Services on OpenShift
  2. OSPRH-11035

ML2/OVN minimum bandwidth QoS (backend enforcement)

XMLWordPrintable

    • ML2/OVN minimum bandwidth QoS (backend enforcement)
    • False
    • Hide

      None

      Show
      None
    • False
    • RHOSSTRAT-563Feature - ML2/OVN minimum bandwidth QoS (backend enforcement)
    • Not Selected
    • Proposed
    • Proposed
    • Done
    • RHOSSTRAT-563 - Feature - ML2/OVN minimum bandwidth QoS (backend enforcement)
    • openstack-neutron-22.2.2-18.0.20250609044743.51e19bb.el9osttrunk
    • Proposed
    • Proposed
    • 0% To Do, 0% In Progress, 100% Done
    • Neutron Sprint 1, Neutron Sprint 2, Neutron Sprint 3, Neutron Sprint 4, Neutron Sprint 5, Neutron Sprint 6, Neutron Sprint 8, Neutron Sprint 9, Neutron Sprint 10, Neutron Sprint 11, Neutron Sprint 12, Neutron Sprint 13, Neutron Sprint 14, Neutron Sprint 15, Neutron Sprint 16, Neutron Sprint 17
    • 16

      This epic cover the effort to implement the ML2/OVN minimum bandwidth QoS support.

      During the testing of [1], it was found that Neutron ML2/OVN QoS extension is using two OVN QoS APIs:

      • The OVN QoS registers. These are used for DSCP marking and bandwidth policing (not shaping). It was enough when the minimum bandwidth rule was not supported.
      • The LSP.options rate values [2], that allow to define maximum and minimum bandwidth.

      In order to define a minimum bandwidth rule, it is also needed to define a maximum rule in the LSP.options dictionary. The previously implemented maximum bandwidth feature is no longer valid and needs to be refactored.

      This issue is affecting to all existing versions:

      • master U/S, that will be the future OSP18 [3]
      • OSP18 (this epic)
      • OSP17 (won't be supported, still under decision)

      DoD:

      • Implementation of the code in U/S and backported to OSP18.
      • Testing, done by QA team.
      • Documentation.

      NOTE: there is a missing feature in core OVN that is tracked in [4]. It is needed to set the HTB max rate for the parent class ID of the TC qdisc created in the egress interface. In case of not having a value, the default rate will be 10Gbps. That usually happens in the virtualized environments: the virtual NICs don't provide a speed link value. This is why https://issues.redhat.com/browse/OSPRH-11046 provides info about how to set this value during testing.

      [1]https://bugzilla.redhat.com/show_bug.cgi?id=2077681

      [2]https://www.ovn.org/support/dist-docs/ovn-nb.5.html

      [3]https://bugs.launchpad.net/neutron/+bug/2015376

      [4]https://issues.redhat.com/browse/FDP-951

              rodolfo_alonso Rodolfo Alonso
              rodolfo_alonso Rodolfo Alonso
              Bharath M V Bharath M V
              rhos-dfg-networking-squad-neutron
              Votes:
              0 Vote for this issue
              Watchers:
              5 Start watching this issue

                Created:
                Updated:
                Resolved: