Uploaded image for project: 'Migration Toolkit for Virtualization'
  1. Migration Toolkit for Virtualization
  2. MTV-1867

[Scale] Efforts related to multi-network selection troubleshooting in scalelab

XMLWordPrintable

    • Icon: Task Task
    • Resolution: Unresolved
    • Icon: Undefined Undefined
    • None
    • None
    • Scale&Perf-QE
    • 5
    • False
    • None
    • False
    • Hide

      Currently the ability to ping publicvlan outside the lab shows the issue is a configuration issue related to the deployment and not dns related.

      Show
      Currently the ability to ping publicvlan outside the lab shows the issue is a configuration issue related to the deployment and not dns related.

      Task:

      Deploy Cloud 20 as MNO deployment via jetlag with public vlans enabled.

      Cloud 20 publicvlan should use

      vlanid IP range Netmask Gateway
      603     10.1.52.0/24 255.255.255.0 10.1.52.254

       

      Background
      We are seeing some strange behavior across our clouds related to public vlan assignment.
      The following clouds are assigned to the corresponding vlan.
      Cloud 33 vlan608
      Cloud 15 vlan611
      Cloud20 vlan603
       
      Note that each of the above clouds 33, Cloud 15, Cloud 20 each have workers which have an interface which is a vlan member of Cloud32, and this vlan exception is desired for these workers to allow for communication from and to cloud 32.
       
      In tickets MTV-1682 dev has cited that we are unable to perform show cert command between clusters meaning the show cert command only works from the provisioner of the cloud that you are currently in, when running the openssl client command in debug we see the connection is dropped immediately for attempting cross cloud (prvisioner of one cloud talking to OCP api of another).
       
      Going through our support cases in the perflab chats and this is a reproduction of the existing issues we've had all along with these environments.  See the original thread
      The concern is that jetski is to blame in firewall configuration and with the changes from openshiftsdn to ovn and from jetski are making this harder to debug,  in order to reduce troubleshooting time, we likely need to have fresh deployment, meaning  I think the best way forward would be to redeploy Cloud20 with jetlag with publicvlan enabled and see if the issues are removed.  This way it will deploy with ovn and not come from an upgrade change, and we'll be certain that we aren't missing deployment related changes.  I suggest cloud20 as it wont block work for cloud33/15. 

       

       

              dvaanunu@redhat.com David Vaanunu
              mlehrer@redhat.com Mordechai Lehrer
              Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

                Created:
                Updated: