Forum Discussion

John_Meggers's avatar
John_Meggers
Icon for Nimbostratus rankNimbostratus
Feb 20, 2013

Node Failover

One of my co-workers is getting ready to do an installation of an appliance-based product that uses an active / standby redudant architecture (instead of load-balanced to active / active). I know I can prioritiize one node over the other using priority group activation, and all traffic will got to node 1 until node 1 fails, and then go to node 2. That's simple.

 

What I can't figure out is how to KEEP traffic on node 2 even after node 1 has become available again. With PGA, once node 1 is recognized as available, LTM will start sending new traffic to him. I'm not sure why it's important for all traffic to remain on node 2 until manual fail-back, but that's the way my co-worker says he wants it to work. Any suggestions on how to make this work? So far the only thing I can think of is to tell him if node 1 fails, go in and manually disable node 1 until it's time to do a manual fail-back.

 

Thanks for any help.

 

John Meggers

 

config
design

2 Replies

Sort By
  • What_Lies_Bene1's avatar
    What_Lies_Bene1
    Icon for Cirrostratus rankCirrostratus
    Feb 20, 2013
    If you configure manual resume on the health monitor applied to the pool, node 1 will stay marked down until you manually enable it. Of course, if you do this you run the risk of node 2 going down before anyone has had a chance to resume a now healthy node 1.

     

     

    This could all be moot of course. If the appliances have a failover mechanism, why not just have a single pool member pointing at whatever the failover shared IP is on the appliance 'cluster' and letting it decide?
  • R_Marc's avatar
    R_Marc
    Icon for Nimbostratus rankNimbostratus
    Oct 31, 2014

    Well, I have similar requirements and I use iControl (Rest) scripts for this. On an event, an alert goes out to the NOC that crap has just gone down, and they run an script to identify the new "master" which just swaps the priorities and then re-enables the bad node (then they can restart, or whatever, the backend that was bad. It could probably also be done with iCall on a monitor down event to reorder the priorities via a script.

     

    In my case, the NOC also has scripts to "normalize" traffic to the original/desired state.