Google respond differently to two identical nginx setups and 200 codes; any ideas?

Question

I'm rather confused...

I have a linode.com VPS which has been cloned recently, so the settings are the same between nginx servers.

One lives on a dev subdomain, one on a www.

I'm trying to run a google experiment on my live server, which claims:

Web server rejects utm_expid. Your server doesn't support added query arguments in URLs.

My logs show on the dev server where it works:

74.125.186.32 - - [13/Sep/2012:13:33:45 -0700] 
    "GET /product/iphone-case/?utm_expid=25706866-0
    HTTP/1.1" 200 12521 "-" "Google_Analytics_Content_Experiments 
74.125.186.32 - - [13/Sep/2012:13:33:45 -0700] 
    "GET /product/iphone-case/?ab_reviews=True&utm_expid=25706866-0
     HTTP/1.1" 200 14679 "-" "Google_Analytics_Content_Experiments

My production server shows google making a second request.

74.125.186.41 - - [13/Sep/2012:13:34:49 -0700] 
    "GET /product/iphone-case/?ab_reviews=on&utm_expid=25706866-1 
    HTTP/1.1" 200 12104 "-" "Google_Analytics_Content_Experiments
74.125.186.41 - - [13/Sep/2012:13:34:49 -0700] 
    "GET /product/iphone-case/?utm_expid=25706866-1 
    HTTP/1.1" 200 12122 "-" "Google_Analytics_Content_Experiments
74.125.186.41 - - [13/Sep/2012:13:34:49 -0700] 
    "GET /product/iphone-case/ <--- A second request for some reason. 
    HTTP/1.1" 200 12522 "-" "Google_Analytics_Content_Experiments

I'm not sure how google determines why it needs to send a second request without the querystring. The original request has clearly sent a 200 OK status response.

Does anybody have any suggestions where to look next? The HTML (compared by diff) on the two pages is exactly the same.

I would be tempted to fire up wireshark and try to get a full capture of the requests/replies. — Zoredache, Sep 13 '12 at 20:46

score 0 · Accepted Answer · answered Sep 13 '12 at 21:19

I'm not sure what caused this; but it has been fixed now.

For anyone with similar problems, I can only suggest looking into the HTML document. I think my issue may have had to do with running a test on a development environment (dev.foo) and setting up a campaign as such; then moving that campaign to a live environment with the same tracking code.

Throw in a dash of issues with caching various versions of these pages and we have my problem.

I'm sorry I can't be clearer about it!

Google respond differently to two identical nginx setups and 200 codes; any ideas?

1 Answers1