Uploaded image for project: 'Fast Datapath Product'
  1. Fast Datapath Product
  2. FDP-586

[RFE] Active/Passive replication should be done with update2 and not update

    • Icon: Epic Epic
    • Resolution: Unresolved
    • Icon: Major Major
    • None
    • None
    • ovs-dpdk
    • [RFE] Active/Passive replication should be done with update2 and not update
    • 21
    • False
    • False
    • Hide

      Please mark each item below with ( / ) if completed or ( x ) if incomplete:

      ( ) The acceptance criteria defined below are met.

      Given an OVS/OVN deployment using active/passive replication with many database objects,

      When replication occurs due to changes in the database,

      Then ovsdb-server uses update2 instead of update so that the network traffic and CPU load associated with replication is reduced.


      ( ) The epics work is available in a downstream build (nightly/async or other)


      ( ) Test coverage is available in downstream CI if applicable


      ( ) All cards under the epic have been moved to Done


      ( ) Failed Test Plans have bugs added as children to the epic/feature.

      Show
      Please mark each item below with ( / ) if completed or ( x ) if incomplete: ( ) The acceptance criteria defined below are met. Given an OVS/OVN deployment using active/passive replication with many database objects, When replication occurs due to changes in the database, Then ovsdb-server uses update2 instead of update so that the network traffic and CPU load associated with replication is reduced. ( ) The epics work is available in a downstream build (nightly/async or other) ( ) Test coverage is available in downstream CI if applicable ( ) All cards under the epic have been moved to Done ( ) Failed Test Plans have bugs added as children to the epic/feature.
    • rhel-9
    • rhel-net-ovs-dpdk
    • 100% To Do, 0% In Progress, 0% Done
    • Undefined
    • ssg_networking

      This epic tracks all the effort needed to deliver the solution related to the feature request described below.
      Description of problem:

      When using active/passive replication, ovsdb/replication.c uses "update" notifications, which means that the full change is sent across the wire instead of just the diffs. This is especially noticeable in the case of the "drop" port group (RFE for feature to allow removal here https://bugzilla.redhat.com/show_bug.cgi?id=1947807) and with Logical_Switches that have a lot of ports. In the former case, I've see 17000 entries in that port group that are sent during every single port create.

      Given the load we see on ovsdb-server, it seems like not having to iterate all of those objects to create/process those notifications would be helpful.

      Version-Release number of selected component (if applicable):
      FDP 21.C

      How reproducible:
      100%

      Steps to Reproduce:
      1. Set up active passive replication
      2. Watch the update notifications between servers
      3.

      Actual results:
      "update" notifications are sent

      Expected results:
      "update2" notifications are sent

      Additional info:

              ovnteam@redhat.com OVN Team
              twilson@redhat.com Terry Wilson
              Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

                Created:
                Updated: