Forum Discussion

mwinstead_13706's avatar
mwinstead_13706
Icon for Nimbostratus rankNimbostratus
Dec 23, 2013

Round Robin Load Balancing

I have a two node WEB pool with the virtual IP set to dest_address persistence and the Pool is using Round Robin as its load balancing logic.

 

Every time we hit the WEB URL we are directed to the same server every time. If I disable that pool member we hit the other member fine and the site works fine so I know the pool members are both functioning. Seems like it's just not load balancing properly. I had about 10 different users (Most haven't gone to the URL before) hit the web URL and they were all directed to the same server each time. That cant be correct behavior.

 

If I remove one member and then add it back then everyone hits only the other member and vice versa if I remove the other member and add it back.

 

Any thoughts/ideas on this ?

 

FYI - Running BIG-IP v11.4.1 (Build 625.5)

 

Thanks,

 

APM
availability
design
devops
iApps
management
security

4 Replies

Sort By
  • nathe's avatar
    nathe
    Icon for Cirrocumulus rankCirrocumulus
    Dec 23, 2013

    Are you meaning to use dst-address persistence? this will always send you to the same pool member based on the dst address in the packet. So if the dst address is the same then you'll see the symptons you describe.

     

    HTH

     

    N

     

  • nathe's avatar
    nathe
    Icon for Cirrocumulus rankCirrocumulus
    Dec 23, 2013

    So that's telling me the destination address is 10.10.10.16 the virtual is 10.10.10.10 and the persistence is to pool member 10.10.10.14. So any connections to the same dst will go to the .14 pool member.

     

    Do you not require source add persistence instead?

     

  • nathe's avatar
    nathe
    Icon for Cirrocumulus rankCirrocumulus
    Dec 23, 2013

    Add all-properties after the tmsh cmd and it should give you more detail.

     

  • nathe's avatar
    nathe
    Icon for Cirrocumulus rankCirrocumulus
    Dec 23, 2013

    The persistence record may have timed out. Looking at the stats you're always going to get the symptoms you describe because there's only one DST add ie the virtual server address. You would normally have DST add persistence with a wildcard vs where the DST could be different. I suggest you remove the persistence type from dst to source.

     

    Worth noting that persistence checks happen prior to load balancing so if there's a persistence record then the load balancing algorithm isn't referenced.