cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
374
Views
0
Helpful
1
Replies

CSS11501 sticky not remapping existing flows

rossd
Level 1
Level 1

Hi

I have two servers and a single VIP for a website which requires stickiness. I use a specific uri for keepalives so the servers can be taken from service if the keepalive uri becomes unavailable.

The problem I have is that if I rename the keepalive file on the server resulting in the keepalives going to "Down", existing client connections arent remapped for up to 60 seconds afterwards.

New connections appear fine and are routed to the functioning server.

Config

service server1

ip address 10.12.1.244

keepalive type http

keepalive port 80

protocol tcp

keepalive method get

port 80 range 3001

keepalive uri "/alive.htm"

active

service server2

ip address 10.12.1.245

keepalive type http

keepalive port 80

protocol tcp

keepalive method get

port 80 range 3001

keepalive uri "/alive.htm"

active

content RG-PROD-1

vip address 10.123.190.20

add service server1

add service server2

protocol tcp

no persistent

advanced-balance sticky-srcip

active

For example, if as a client I am opening http://10.123.190.20/alive.htm in my browser and I then rename the alive.htm on server1 causing the keepalive->Down, I just sit there hitting refresh for up to 60 seconds and during that time I keep getting the expected 404 not found from the server until some sort of timeout kicks in and I get correctly moved to the functioning server.

Am I missing something obvious here? I use CSM's as well and have no such problems.

Thoughts appreciated.

Ross

1 Reply 1

rossd
Level 1
Level 1

Resolved.

It took a lot of searching but identified via post http://forum.cisco.com/eforum/servlet/NetProf?page=netprof&forum=Data%20Center&topic=Application%20Networking&topicID=.ee7814f&fromOutline=&CommCmd=MB%3Fcmd%3Ddisplay_location%26location%3D.ee94177

Essentially, the client does not get remapped unless it receives a RST from the balancer when the service goes down. This functionality only works if you are close enough to have an arp for the device or you have the content rule configured to include the line: url "/*". Making it a L5 rule(not sure what you can do about raw tcp socket connections).

This resolved the issue. Please see the above post for a better explanation.

Much kudos to Gilles for his earlier post.

Cheers

Ross

Review Cisco Networking for a $25 gift card