Cisco ACE 4710 - Health Monitoring for Real Servers

Unanswered Question
Oct 1st, 2010

Hi,

I have setup the following health probe to check for the existence of a specific web page.  My intention is that when the web page is removed, the health check fails and the rserver status changes to 'out of service'.  Unfortunately, when I remove the web page, I see the health check fail, and the rserver state change to 'PROBE-FAILED', however the rserver does not go 'out of service' and continues to respond to requests.

Can anyone see where I'am going wrong?

Health check probe config

-------------------------------------

probe http live_http_int
  interval 15
  passdetect interval 60
  request method get url /loadbalancer/internal.html
  expect status 199 201
  open 10

RSERVER config

--------------------------

rserver host Server1
  description Server1
  ip address 10.10.10.1
  conn-limit max 4000000 min 4000000
  probe live_http_int
  inservice
rserver host Server2
  ip address 10.10.10.2
  conn-limit max 4000000 min 4000000
  probe live_http_int
  inservice

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
louis-philippe.lison Fri, 10/01/2010 - 09:51

Hi syannetwork,

I think you have to "force" the failed server to close the connection when it has failed. Otherwise it will still serve the available HTML pages.

Have a look at the "Configuring the ACE Action when a Server Fails" in the "Cisco Application Control Engine Module Server Load-Balancing Configuration Guide" and let me know if the following command helped:

conf t

serverfarm host ServerFarm

failaction purge

Have a good WE.

Cheers

LPL

Daniel Arrondo Ostiz Fri, 10/15/2010 - 02:38

Hi sysannetwork,

On the ACE, when a probe goes down, the server will stop to receive new connections, but the currently established ones will remain.This is to allow connections to continue working if by any chance the server is still able to reply (for example, if the web page queried by the probes is different than the one requested by the client)

This behavior can be changed with the "failaction" command suggested by louis-philippe.lison, you can define "failaction purge" to immediately close all connections or "failaction reassign" to move the connections to a different rserver in the backend transparently to the client.

Actions

This Discussion