multi server addresses appear in variable $upstream_addr which was not supposed to

[褚夫元]

Hello!

    It is pretty strange that there was two upstream servers in $upstream_addr. It seems two servers received request and both had a respose. Client received the response which took less duration time, I guess.  Am I right ?

    What is stranger is  the name of upstream server group appears in $upstream_addr. I don't know the consequence in such situation. Any explain ?

    *More important is how to find the "spirit" behind this weird problem and fix it.* 

    Thanks a lot ! !

    Nginx Version: nginx/1.6.2
    OS:  CentOS release 6.5 (Final) x86_64

    Here is access log below:

client_ip      26/Aug/2016:15:45:06 +0800      GET     host_ip       80      10.10.10.10:8080, backendservice       /backend/api/v1.0/bowner/1123223      HTTP/1.1        502     551     -       UA:Java/1.8.0_60        -       0.139   0.139, 0.000

client_ip      26/Aug/2016:15:45:12 +0800      GET     host_ip       80      10.10.10.10:8080, backendservice       /backend/api/v1.0/bowner/1123822      HTTP/1.1        502     551     -       UA:Java/1.8.0_60        -       0.070   0.070, 0.000

client_ip      26/Aug/2016:15:45:17 +0800      POST    host_ip       80      10.10.10.10:8080, 10.10.10.11:8080        /backend/api/v1.0/bowner/1133782
       HTTP/1.1        200     129     -       UA:Java/1.8.0_60        -       0.124   0.043, 0.081


* Backend API  /backend/api/v1.0/bowner/{request_id} is supposed to be processed by one server. One server will be return 500 code if many backend servers received same request at (almost) same time.

Here is upstream configuration:
 
upstream backendservice {
        server 10.10.10.10:8080weight=15;
        server 10.10.10.11:8080 weight=13;
}





-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.nginx.org/pipermail/nginx/attachments/20160828/767dad6e/attachment.html>