Failover issues with sticky

Hello,

I’m using nginx solely as a load balancer, with nginx-sticky (
Google Code Archive - Long-term storage for Google Code Project Hosting.) as a way to maintain
sticky
sessions.

There are 2 upstream servers configured - and I have to occasionally
mark
one of them as down. When I do this, failover from clients that have a
sticky session cookie does not seem to happen to the other upstream
server
(which is up).

Here are the steps in detail -

  1. nginx with 2 upstream servers in the backend s1, s2
  2. Client connects, nginx routes to s1
  3. s1 marked as down in config, nginx config is reloaded
  4. Requests from the same client continue going to s1, instead of s2

My understanding is that nginx should automatically route traffic to the
second server. Anything I am missing here?

  • Hrish

2011/7/21 Hrishikesh B. [email protected]:

Here are the steps in detail -

  1. nginx with 2 upstream servers in the backend s1, s2
  2. Client connects, nginx routes to s1
  3. s1 marked as down in config, nginx config is reloaded
  4. Requests from the same client continue going to s1, instead of s2

My understanding is that nginx should automatically route traffic to the
second server. Anything I am missing here?

Hello

it’s a bug on the sticky module. Can you open a bug report there
(Google Code Archive - Long-term storage for Google Code Project Hosting.) please ?

I’ll look at it.

++ jerome

Hi Jerome,

Done - Google Code Archive - Long-term storage for Google Code Project Hosting.

  • Hrish

2011/7/21 Jrme L. [email protected]

Thanks a lot for the quick turnaround, Jerome.
I’ll test it as soon as I get a chance.

Hrish

2011/7/21 Jrme L. [email protected]

the bug has been fixed. I’ve released the 1.0-rc3 for the opportunity:

http://code.google.com/p/nginx-sticky-module/downloads/list

Can you please test it ?

++ jerome

2011/7/21 Hrishikesh B. [email protected]: