Forum Discussion

mwi's avatar
mwi
Icon for Cirrus rankCirrus
Nov 15, 2023

F5 LTM manual resume problem

Hi,

we have configured a pool with two nodes, one primary and one secondary, every node has its own monitor.

The primary node monitor has set the option "manual resume" so if this node isnt avalible, the traffic goes to secondary, but doesnt switch automaticly back.

It works under normal circumstances fine, but yesterday we had some network issues so that the active loadbalancer lost the connection to the primary node, the passive loadbalancer had no problems.

We had to take down the active Loadbalancer and now the problem was that the before passive loadbalancer routed traffic to the primary node again.

Is there any solution to sync the node status automaticly?

 

Best regards.

 

  • mwi 
    This is normal because it needed to be manually resume after being disables.

    you need to implement another approach of "Priority Group Activation"
    That's fit your Solution.

    K13525153: Configure a standby pool member to process traffic when primary pool member goes down
    Refer to this and you would get it all -> https://my.f5.com/manage/s/article/K13525153

  • mwi 

    So if i understand your issue is fail back not fail over.

    What i believe is happening is that when you bring the "active" node back online (so both are now green) traffic is still on the "backup" node.  Now i believe the behaviour here is that the active node should start taking on new traffic, but existing traffic will stay on your existing live node so in this case the backup node until the connection is broken or timed out.
    If you want the backup node to stop processing traffic, you need to find a clean way of closing the connections so they are remade at the f5 from the client.  This will depend on your application but this might not be too graceful.

    You could try putting the backup node into disabled or offline modes to help with this process but if the application doesn't open and close ports regularily this may not help you. So you'll be back to working out a way that you can close the connections on the backup server so they are remade on the active server.
    It would be worth testing this with your application if you have that option.

  • M_Saeedthis is allready configured and works. Sorry i didnt mention it.

    The Problem is if the active part of the HA-Cluster lost the connection to the primary node, and the passive part doesnt, after a failover, the primary node is automaticly active again.

    I have drawn a little picture, i hope this makes more clear what my problem is.

     

    PSFletchTheTeki think thats not part of my problem.

    • M_Saeed's avatar
      M_Saeed
      Icon for Cirrus rankCirrus

      mwi 
      Let's assume you would seperate both into two different pools and we would conditionallay govern this via an iRule.


      when CLIENT_ACCEPTED {
      set default_x_pool [LB::server pool]
      if {[active_members $default_pool] < 1}{
      pool Service_backup_y_pool
      }
      }

      • mwi's avatar
        mwi
        Icon for Cirrus rankCirrus

        M_Saeedgood idea, but the primary node would be active on the second HA-Cluster member.

        Did you see the Picture?

  • you can add Gateway Pool into HA setting.
    so if Node's GW pool status is down, it wont be Active node.