CSM: HTTP probe to check health of a farm with real servers on different ports

[sgonsalv]

Hi Guys,

We have a farm and probe setup as follows:

probe ESS-QAT http
request method get url /irj/portal
  expect status 200
  interval 15
  failed 5
  open 5
  receive 5
!
serverfarm ESS-QAT
  nat server
  no nat client
  predictor leastconns
  failaction purge
   real 130.194.27.226  50100
   inserv
   real 130.194.27.227  51000
   inserv
  probe ESS-QAT
!

vserver ESS-QAT-DECVIP

description decrypt VIP for ESS

  virtual 172.16.11.115 tcp www
  serverfarm ESS-QAT
  persistent rebalance
  parse-length 4000
  sticky 60 group 181
  inservice
!

My question has to do with the probe used to check the health of the servers, and more so clarify my understanding of the probes behavior.  You'll notice that the http probe doesn't have a port specified.  To my understanding (based on reading the config guide), the HTTP probe will inheret (and therefore use) the ports attached to each of the real servers when performing the GET.  That is perform a GET on 130.194.27.226:50100 and 130.194.27.227:51000 respectively.  Is this correct?

When testing this, i noticed that the probe fails on both servers...i'm hoping i don't need to configure a TCL script to check the health of the servers given this configuration.

Note, usually we have both servers recieving connections on the same port number...however in this situation each server is recieving connections on different ports.

thanks

Sheldon

[jsirstin]

Sheldon,

Your understanding of the probe port inheritance is correct. The CSM should be sending the probe to each of the two servers on different ports with your config. Would it be possible to get a sniffer capture of the server reply to confirm if it is giving a 200 ok response?

[sgonsalv]

Hi,

Thanks for the response to clarify my understanding - i'll have a closer look at the traffic flow.

Cheers

Sheldon