Uploaded image for project: 'Fast Datapath Product'
  1. Fast Datapath Product
  2. FDP-770

OVN Kubernetes multi-homing in CNV: Seamless live-migration

XMLWordPrintable

    • Icon: Task Task
    • Resolution: Done
    • Icon: Major Major
    • None
    • None
    • OVN
    • 3
    • False
    • Hide

      None

      Show
      None
    • False
    • Hide

      Given a KubeVirt VM running on nodeA in a Kubernetes cluster, where OVN supports multiple chassis,

      When a live migration of the VM is initiated from nodeA to nodeB using a transit switch setup,

      Then, during the migration, the requested-chassis parameter should:

      • Be set to nodeA when only the source pod (VM on nodeA) exists,
      • Be updated to nodeA,nodeB when both source and destination pods (VMs on nodeA and nodeB) exist simultaneously during migration,
      • Be set to nodeB once the migration completes and only the destination pod (VM on nodeB) remains.

      If any issues are seen during the testing, a Jira ticket must be filed with detailed findings. 

      Show
      Given a KubeVirt VM running on nodeA in a Kubernetes cluster, where OVN supports multiple chassis, When a live migration of the VM is initiated from nodeA to nodeB using a transit switch setup, Then, during the migration, the requested-chassis parameter should: Be set to nodeA when only the source pod (VM on nodeA) exists, Be updated to nodeA,nodeB when both source and destination pods (VMs on nodeA and nodeB) exist simultaneously during migration, Be set to nodeB once the migration completes and only the destination pod (VM on nodeB) remains. If any issues are seen during the testing, a Jira ticket must be filed with detailed findings. 
    • OCPSTRAT-1613 - Enhance User Defined Networks: Add support for Services(Full), EIP(Full), NetPol: GA
    • FDP 24.H

      Goal

      Ensure the multi requested chassis feature can be used on an IC cluster.

      We want to use this for KubeVirt VM live-migrations. 

      KubeVirt VMs run in a Kubernetes pod, thus, we are planning on doing something like:

      (assuming the VM is scheduled in nodeA and will be migrated to nodeB)

      t=0: (only src pod exists) requested-chassis = nodeA
      t=X: (mid migration; src and dst pod exist) requested-chassis = nodeA,nodeB
      t=y: (migration has ended; only dst pod exists) requested-chassis = nodeB

      Scope of this issue

      • Make sure that the requested-chassis parameter can handle migrations in environments where nodes are connected to different logical switches.
      • Test that multiple-chassis and the activation-strategy work effectively for transit switches in OVN Kubernetes.
      • If there are issues, create Jira tickets.

      The issue only involves testing migration from one node to another when the pod or VM is directly attached to a transit switch.

      Notes

      • Utilize requested-chassis parameter
      • Ask Ales Musil for assistance
      • Even though OVN should provide support for seamless live-migration, there may be issues: https://bugzilla.redhat.com/show_bug.cgi?id=1903653
      • This is already done in Neutron and Kube OVN. See the presentation from KubeVirt Summit 2024

      Done Checklist

      Who What Reference
      DEV Upstream roadmap issue <link to GitHub Issue>
      DEV Upstream code and tests merged <link to meaningful PR>
      DEV Upstream documentation merged <link to meaningful PR>
      DEV gap doc updated <name sheet and cell>
      DEV Upgrade consideration <link to upgrade-related test or design doc>
      DEV CEE/PX summary presentation label epic with cee-training and add a <link to your support-facing preso>
      QE Test plans in Polarion <link or reference to Polarion>
      QE Automated tests merged <link or reference to automated tests>
      DOC Downstream documentation merged <link to meaningful PR>

       

              amusil@redhat.com Ales Musil
              mduarted@redhat.com Miguel Duarte de Mora Barroso
              Jianlin Shi Jianlin Shi
              Votes:
              0 Vote for this issue
              Watchers:
              7 Start watching this issue

                Created:
                Updated:
                Resolved: