Uploaded image for project: 'Network Observability'
  1. Network Observability
  2. NETOBSERV-1344

New "Inner" flows are excluded from metrics

Details

    • Bug
    • Resolution: Done-Errata
    • Major
    • netobserv-1.4.1
    • netobserv-1.4
    • FLP, Operator
    • None
    • False
    • None
    • False
    • Hide
      In 1.4, the "Inner" flow direction was introduced to account for flows between pods running on the same node. Flows with the "Inner" direction were not taken into account in the generated Prometheus metrics derived from flows, resulting in under-evaluated bytes and packets rates.
      Now, derived metrics are including flows with the "Inner" direction, thus providing correct bytes and packets rates.
      Show
      In 1.4, the "Inner" flow direction was introduced to account for flows between pods running on the same node. Flows with the "Inner" direction were not taken into account in the generated Prometheus metrics derived from flows, resulting in under-evaluated bytes and packets rates. Now, derived metrics are including flows with the "Inner" direction, thus providing correct bytes and packets rates.
    • NetObserv - Sprint 243

    Description

      Current metrics are using a filter for deduplication, such as: "FlowDirection: 0"

      Which makes the flows having the new Inner state being excluded from the metrics, which they shouldn't.

      This results in under-evaluated traffic rates in metrics

       

      Reproduction steps:

      • Deploy default netobserv w/ sampling=1
      • Deploy some workload providing consistent network load, for instance mesh-arena:
      • Let it run ~5 minutes
      • Check netobserv plugin, Overview tab, filter on Src ns=mesh-arena => you should get a byte rate ~31 KBps
      • Check Observe > Metrics, enter promql:
        • sum(rate(netobserv_workload_ingress_bytes_total{SrcK8S_Namespace="mesh-arena"}[1m]))

      Expected results:

      Same as in overview tab, ~31KBps

      Actual results:

      It's less than expected, around 23 kBps (in fact, that will depend on how pods are allocated to nodes)

      Attachments

        Activity

          People

            jtakvori Joel Takvorian
            jtakvori Joel Takvorian
            Nathan Weinberg Nathan Weinberg
            Votes:
            0 Vote for this issue
            Watchers:
            5 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: