Forum Discussion

Lightspeed_VT_5's avatar
Lightspeed_VT_5
Icon for Nimbostratus rankNimbostratus
Jul 18, 2008

Redundant pair and switch fault tolerance

Sorry if this hardware question is off-topic, but didn't see anywhere else to post. Also, I'm a developer stuck with a network implementation, so bear with me on that, too.

 

 

We have two LTM 1500s that are going to run as a standby system in a redundant pair using the serial cable.

 

 

We are currently running the application LAN on two fault tolerant switches that auto failover if there is a problem.

 

 

I'm wondering how to hookup the LTMs to the switches. Do I run two interfaces from each LTM--one to each switch? The active switch would send traffic and the standby switch would obviously send nothing.

 

 

I'm sure this is a common network config. Can someone shed some light on this?
config
design

10 Replies

Sort By
  • hooleylist's avatar
    hooleylist
    Icon for Cirrostratus rankCirrostratus
    Jul 21, 2008
    Yes, you could connect each LTM to both switches on the same VLAN. Assuming both the LTMs and the switches are in active-passive pairs, this should work fine. Any network gurus have suggestions?

     

     

    Aaron
  • JRahm's avatar
    JRahm
    Icon for Admin rankAdmin
    Jul 21, 2008
    I'd need a greater discovery of the environment to weigh in with an informed opinion. Blind, I'd recommend connecting the first LTM to the first switch, and the second LTM to the second switch. For the greatest redundancy, you could connect like this: 
      
      
   SWITCH1                                     SWITCH2   
        ||                                                  ||                
        ||                                                  ||    Client-side LACP Trunks (client vlans)   
        ||                                                  ||     
   LTM 1500 =================LTM1500   
        ||       (Mirror/Sync LACP Trunk)        ||     
        ||                                                  ||    Server-side LACP Trunks (server vlans)   
        ||                                                  ||    
   SWITCH1                                     SWITCH2

    If you have multiple blades in each chassis, you can connect each leg of an LACP trunk into each blade for even greater redundancy. Make sure you check the asic limitations on your switch linecards before assigning ports, most cards oversubscribe the ports so you want to be as optimal as possible in your assignment.
  • JRahm's avatar
    JRahm
    Icon for Admin rankAdmin
    Jul 21, 2008
    This assumes you aggregate two physical cables into one logical connection. Cisco calls it etherchannel. The standards-based technology is LACP trunking. If you don't have the Fiber SFP's installed and don't plan on buying them, you don't have enough ports for this approach and you'll need to do a single cable for each leg (clientside/serverside/box-to-box)

     

     

  • Lightspeed_VT_5's avatar
    Lightspeed_VT_5
    Icon for Nimbostratus rankNimbostratus
    Jul 21, 2008
    Thanks for the diagram. There are a couple things I don't understand: 1) how to aggregate ports on the LTM. Is there a way to do this in the GUI, or is it a command-line thing and 2) Why do both aggregate links go to the same switch?
  • JRahm's avatar
    JRahm
    Icon for Admin rankAdmin
    Jul 21, 2008
    1) Network->Trunks->Create.

     

    2) I'm not sure F5 supports multi-chassis etherchannel yet (or on the switch side, whether the virtual switching system is installed). Spanning-tree is most likely a requirement in any scenario short of VSS, and I prefer to avoid that on the LTM if possible.
  • JRahm's avatar
    JRahm
    Icon for Admin rankAdmin
    Jul 21, 2008
    Re: 2, most design criteria attempts to eliminate single points of failure. By aggregating to one switch, you are protected if one cable fails, if one line card fails, etc, but if the switch fails, well, that's what your other LTM and switch are for!

     

     

    The LTM itself is a switch, and treating it as such when not a requirement adds unnecessary complication to the layer2 architecture.
  • zafer's avatar
    zafer
    Icon for Nimbostratus rankNimbostratus
    Jul 27, 2008
    i want add some questions to this diagram,

     

     

    if i want connect bigip units cross to Cisco 6500; it means new topology would be like this;

     

     

     

    bigip 1; vlanx (1.1 and 1.2); 1.1 connected cisco1 and 1.2 connected cisco2

     

    bigip 2; vlanx (1.1 and 1.2); 1.1 connected cisco1 and 1.2 connected cisco2

     

     

    and bigip1-2 directly connected

     

     

    rstp enabled on cisco and cisco blocks 1.2 interfaces path

     

    the problem is cisco guys traffic on 1.2 ports and its not working

     

    it means, client ping vips traffic pass over 1.1 interface but responces goes over 1.2 (blocked on cisco interface)

     

     

    how can i solve this problem?

     

     

    regards

     

     

    zafer