Uploaded image for project: 'Network Observability'
  1. Network Observability
  2. NETOBSERV-1805

netobserv 1.6.1 is creating vast number of threads on OCP 4.14 with errors mentioning tcx binary is not available.

    • Improvement
    • False
    • False
    • Hide
      1- When added the secondary interface support we noticed we need to iterate multiple time to be able to register per netns to netlink to learn about interface notification, while we forgot to free all unsuccessful handler to avoid leaking FD.
      2- When netns was deleted we didn't setup go close channel event to terminate netlink's goroutine socket causing go threads leak.
      Show
      1- When added the secondary interface support we noticed we need to iterate multiple time to be able to register per netns to netlink to learn about interface notification, while we forgot to free all unsuccessful handler to avoid leaking FD. 2- When netns was deleted we didn't setup go close channel event to terminate netlink's goroutine socket causing go threads leak.
    • NetObserv - Sprint 259

      Description of problem:
      netobserv 1.6.1 is creating vast number of threads on OCP 4.14 with errors mentioning tcx binary is not available.

      Steps to Reproduce:

      
      

      1. Install netobserv 1.6.1 on OCP 4.14.z
      2. on high load ebpf pod starts creating vast no. of threads . Mentioning in pods logs that the node don't have the TCx binary available
      3. Pod has to restart or crashes eventually.

      Actual results:
      ebpf pods are creates lots of threads, and need to restart.

      Expected results:
      No issue should be there when the pod is running ebpf agent pod in high load environment.

              mmahmoud@redhat.com Mohamed Mahmoud
              rhn-support-manyayad Mahesh Nyayadhish
              Mehul Modi Mehul Modi
              Votes:
              0 Vote for this issue
              Watchers:
              10 Start watching this issue

                Created:
                Updated:
                Resolved: