Forum Discussion

ey52_389743's avatar
ey52_389743
Icon for Nimbostratus rankNimbostratus
May 02, 2019

IIS Monitoring

Hi,

 

I am new to load balancing technology and F5. Currently in BIG-IP 12.1.2 Build 2.0.276 Hotfix HF2.

 

Our application team had an issue wherein one of the members/servers in a pool went down but clients were still trying to connect to a specific member/server. I manually set it to Force Offline, application team tested it by initiating new connections, and saw that the attempts were still going through the same down member/server. I understand that when Force Offline is used, it allows existing connections to time out and no new connections are accepted. We waited a significant amount of time after the down member/server was set to Force Offline and yet the same connection attempts to the down member/server were seen.

 

The pool is set to use GET / HTTPS/1.1 for Health Monitor with Interval: 5 secs and Timeout: 16 secs. I want to ask the following:

 

How are existing connections treated when a member/server is marked down or inactive without setting it to Force Offline or Disable? Does F5 follow the Timeout setting of the Health Monitor in used before completely clearing all the existing connections and not accept new ones?

 

Is there an improvement that can be done on the Health Monitor above? May be a better version or addition to the existing.

 

Is there an automatic way for a member/server to be set to Force Offline when F5 marks a member/server as down or inactive?

 

Is there an automatic way to clear the current connections after the step above? I stumbled upon this cli command tmsh delete /sys connection ss-server-addr . I do not see the command available may be because of the OS version but this one is, delete sys connection ss-server-addr and wondering if it does the same thing as the former.

 

Best practices for IIS monitoring

 

Thanks!

 

BIG-IP
devops
TMOS

1 Reply

Sort By
  • Vijay_E's avatar
    Vijay_E
    Icon for Cirrus rankCirrus
    May 02, 2019

    Try using OneConnect with /32 netmask and Action on service down set to Reselect.

     

    By default the F5 makes load balancing decision for the very first HTTP request and subsequent requests will be sent to the same pool member that received the very first HTTP request. By using OneConnect, you will be forcing the F5 to make the load balancing decision for each HTTP request. The action on service down provides the F5 with a course of action when the pool member selected is down.