Forum Discussion

MACTEP's avatar
MACTEP
Icon for Altocumulus rankAltocumulus
Mar 21, 2018

Packet loss and huge delays on data interfaces of LTM VM on VMWare

We have a number of lab LTM VMs. Suddenly all of them started experiencing about 10% packet loss and huge delays (up 1000 ms) on traffic interface. Management interfaces are fine. VMWare team swears there were no changes done on the host. I checked host, vSwitch - found nothing. Tried moving to different host, change networks - same. Traffic interfaces have issues, management does not. Deployed new VM from freshly downloaded OVF on totally different VM host - SAME! New VM has no config except management and traffic interfaces. Using untagged, non-trunked interfaces. Issue is experienced even between 2 F5s connected to the same vSwitch with no physical NICs, so issue is not related to the rest of network infrastructure. Swapping management and traffic networks does not help - issue persists only on network connected to traffic interface, so does not seem to be VMWare issue. Looks like it is something wrong with F5 VMs, but it happened to all of them, those that were not touched for years as well as freshly deployed. Has anyone experienced similar issue on VMWare?

 

LTM
Super-NetOps

3 Replies

Sort By
  • Surgeon's avatar
    Surgeon
    Ret. Employee
    Mar 21, 2018

    It is hardly could be big-ip issue. All boxes at the same time even newly deployed. There is a very very little chance, I would say less then 99% that big-ip related issue.

     

    is there any loop in the service vlan? DO you see any traffic increase?

     

    Can you do a packet capture on the big-ip and see how the big-ip process the traffic?

     

  • Surgeon's avatar
    Surgeon
    Ret. Employee
    Mar 21, 2018

    ok, you are taking about delay. The ping you show tell us nothing. Can you run capture on the big-ip against the service.

     

    e.g tcpdump -s0 -vvv -nni 0.0:nnnp -w /var/tmp/$HOSTNAME.pcap host < client_ip>

     

    Access one of the affected vips, get the issue, stop the capture. Show us client and related server side flow using screen shots. include time column in the wireshark. Then you can see where the delay is.

     

  • Surgeon's avatar
    Surgeon
    Ret. Employee
    Mar 23, 2018

    Ilya, even if the issue is with the big-ip we need a packet capture to see where is the issue. Ping tells us nothing.

     

    I am not sure what if it is license related but what is the speed limit in your license? What if your big-ip is dropping the packets due to license speed limit? It might be just legitimate drop of the look packets due to lack of listener.

     

    If the issue affects all boxes in your environment this is definitely something causing this from outside

     

    If you are not facing the issue with ver 12 why just not go with it? What does prevent you to go with 12? There are too many questions and too many unknowns. Need a capture