Forum Discussion

Brian_Kenworthy's avatar
Brian_Kenworthy
Icon for Nimbostratus rankNimbostratus
Feb 01, 2008

Proxy Conns vs Pool Member Conns

Hello experts,

 

 

This has been bugging me for a long time and an issue this morning has prompted me to ask about this...using Bigip 4.5.11, I have a proxy server configured which forwards traffic to a virtual server which is using an irule to split traffic.

 

 

The default pool is an https pool with four members included. Typically, we see approx 300 current proxy connections in the proxy stats tab in the UI, but when in the pool stats tab there will be no more then 5 current connections between all four servers. I am using a least connections (member) LB method with an active http cookie insert for my persistence type on this pool. I am trying to determine why the proxy connections are not appearing as member connections on the pool stats, and if they even should be bas on my configuration? It is difficult to determine the number of current connections to each member without looking directly on the web server itself. The other pool that is specified in the rule appears to accurately show the current member connections in pool stats, although this pool is not using persistence and is round robin.

 

 

If anyone can shed some light on this, it would be greatly appreciated. I can also help to clarify if more info is needed.

 

 

Thanks in advance,

 

Brian
  • Does the connections on the destination host that the proxy server associated match up?

     

     

    /cb

     

  • Hi CB, thanks for the reply.

     

     

    The destination host in my configuration is a virtual server which points to a rule which splits traffic based on URI. If there are certain URI matches, the request is forwarded to an SSL pool with three hosts in it. Everything else is forwarded to a second SSL pool containing four other hosts. If I query the connections on all 7 hosts, would that add up to the number I see in the proxy connections? I am wondering if the one connect settings has any affect on this as well?

     

     

    Also, the client web connections are also proxied at the user end so we may see connection from the same source. Also, on the default pool on the rule, we are using active cookie persistence with insert method.

     

     

    Just trying to make sense of everything...thanks for your help.
  • One connect makes a big difference. You will see different of connections between client and VS and F5 to the pool of servers.
  • I figured. What is the best way to monitor the actual connection counts to the nodes? Any of the SNMP OIDs? It seems it gets difficult when you start throwing proxies in there, oneconnect, etc. We have some perf testing coming up and I wanted to make sure I captured everything.

     

     

    Thanks for the help.
  • Deb_Allen_18's avatar
    Deb_Allen_18
    Historic F5 Account
    Actual connection counts to the nodes the pool members, and protocol profiles (everything that uses the same protocol profile) are all rolled up in SNMP.

     

     

    For connection counts to a specific IP, use node stats:

     

    node_addr ServerCurrConns * .1.3.6.1.4.1.3375.2.2.4.2.3.1.9.*

     

     

    For connection counts to a specific IP+port, use pool member stats:

     

    pool_mem ServerCurrConns * .1.3.6.1.4.1.3375.2.2.5.4.3.1.11.*

     

     

    HTH

     

    /deb