1. Release 1.1 known bugs and limitations

Release 1.1 known bugs and limitations

Home Forums Release Changelogs and Known Bugs Release 1.1 known bugs and limitations

Viewing 3 posts - 1 through 3 (of 3 total)
  • Author
    Posts
  • #1500
    Ilya Baldin
    Participant

      As of Release 1.1 we’ve found a few problems primarily with underlying hardware and software that we are working to fix:

       

      1. Some GPU-equipped worker nodes may randomly reset/reboot themselves when being provisioned. This causes the VM slivers to fail and also affects any other VM slivers already provisioned on the node.

      Resolution: Exact cause is still TBD, we think it relates to the open-source Neuveau NVidia driver on the worker node. The cause has been tracked and is being addressed by updating the driver setup on all GPU nodes.

      Workaround: None

      2. L2STS service may not work between certain sites – the service appears to provision successfully but does not pass traffic.

      Resolution: The problem relates to Cisco IOS software versions on the switches, we believe there is incompatibility between some versions which we are working to address with the vendor We are certain the issue is with the firmware versions on the switches, they are being updated. The firmware has been updated.
      Workaround: None

      • This topic was modified 2 years, 1 month ago by Ilya Baldin.
      • This topic was modified 2 years, 1 month ago by Ilya Baldin.
      • This topic was modified 2 years, 1 month ago by Ilya Baldin.
      • This topic was modified 2 years, 1 month ago by Ilya Baldin.
      • This topic was modified 1 year, 12 months ago by Ilya Baldin.
      #1513
      Ilya Baldin
      Participant

        Issue 1 above is being addressed: Maintenance on FABRIC-MAX and FABRIC-NCSA on Wed 3/16/22 – 5pm ET

        Issue 2 is also being addressed transparently to users by updating IOS firmware on some of the switches.

        #1728
        Xi Yang
        Participant

          3. When site VMs are allocated along with a FABNetv6 service, the IPv6 address on the VM hosts may lose connectivity due to gateway MAC expiration. This was exhibited as ping disruption.

          Resolution: Need to change the timing of FABNetv6 service and host interface activation. Or embed the below workaround in the fablib post-slice action.

          Workaround: Toggle down and up the host interface with the problematic IPv6 address. Then reassign the IPv6 address and re-add the host routes.

           

        Viewing 3 posts - 1 through 3 (of 3 total)
        • You must be logged in to reply to this topic.