Forum Discussion

Marco_Bayarena_'s avatar
Marco_Bayarena_
Icon for Altostratus rankAltostratus
Apr 02, 2008

Failover instead of Load Balancing Pool members?

Is it possible to configure a 2 member pool to have one member receive all the connections until it becomes unavailable then send the traffic to the other member?

 

 

I have two mail servers that I would like to provide failover capabilities. I would prefer to have one mail server to receive mail at one time. When that server fails, I would like the other server to start receiving mail.

 

 

I created a pool that contains both mail servers and a virtual server that points to the pool. None of the load balancing methods of the pool seem to meet my needs. Is there a way to accomplish this?

 

 

Thanks.
config
design

5 Replies

  • Steve_Brown_882's avatar
    Steve_Brown_882
    Historic F5 Account
    Apr 02, 2008
    You need to use priority group activation. Set is up so that it activaties the second group when less than 1 server is availible. The set the server you want to be primary with a priority of 1 and the backup as priority 2. You may be able to find a doc on ask f5 that exsplains it better, but that iis the basic way to do what you are looking to do. I do this for mail servers too and it works well.
  • Daniel_55334's avatar
    Daniel_55334
    Icon for Altostratus rankAltostratus
    Apr 03, 2008
    For this setup, can it be achieved that the backup server remains active after the primary server comes back up?
    • kbose_49650's avatar
      kbose_49650
      Icon for Nimbostratus rankNimbostratus
      Feb 12, 2015
      Perhaps not, because the moment the server with the higher priority comes up it takes precedence. I also think that all existing connections to the existing server (low priority) will be held until they terminate, but new connections will continue to the higher priority server that just came up alive. For what you want, an iRule as described by hoolio
  • hoolio's avatar
    hoolio
    Icon for Cirrostratus rankCirrostratus
    Apr 03, 2008
    There is a nice example of single node load balancing in the iRules Codeshare:

     

     

    Single Node Persistence (Click here)

     

     

     

    The first connection will create a single universal persistence record with a key of "1". All subsequent connections will look up persistence using "1" as the key, resulting in truly universal persistence for all connections. (Use 1 or any constant value. 0 will have the same affect as using 1. One of my customers uses "persist uie [TCP::local_port]"

     

     

    When one node fails, the other is persisted to by all comers. When the 2nd node fails, the 1st again becomes the preferred node for all, ad infinitum.

     

     

    Doesn't offer the capability of manual resume after failure, or true designation of a "primary" and "secondary" instance (sometimes required for db applications), but it sure does solve the problem of "only use one node at a time, I don't care which one, please" (You can use priority to gravitate towards the top of a list...)

     

     

     

     

    Aaron
  • StephanManthey's avatar
    StephanManthey
    Icon for Nacreous rankNacreous
    Feb 12, 2015

    The easiest way to get it done from my perspective works without an iRule.

    Just apply destination address affinity to the virtual server.

    Now check the persistence table: 
    watch -d -n 1 tmsh show ltm persist persist-records

    You will notice a single entry only. Traffic to the virtual server (persistence key) is balanced to selected pool member. No matter who the client is.

    As soon as the pool member fails all incoming traffic will be directed to the alternative pool member and the persistence record will be updated.

    That´s it. Obviously the feature was intended to be used in another context but it works pretty well for this particular purpose.

    I actually started using it this way in BIG-IP v4.2. 🙂

    Thanks, Stephan