3

I have researched this problem, but in most cases the cause of the 502 error is an improperly configured nginx.conf or upstream service. I believe this is different..

As the title suggests, I upgraded ubuntu server 14.04 to 16.04. I use nginx as my web server, and am also running a java/tomcat server, setup in my nginx config as a proxy_pass.

Since the upgrade, every time the server starts up nginx displays error 502: Bad Gateway when attempting to connect to the proxy_pass site. All other sites specified in my config work as expected.

Is it possible that the order in which services are started could cause a persistent 502 error?

To resolve the issue, I must sudo systemctl restart nginx, after which, the proxy_pass service works as expected, until the next reboot.

From the error.log: 

2018/01/24 11:33:20 [error] 1886#1886: *202 connect() failed (111: Connection refused) while connecting to upstream, client: 10.0.0.1, server: localhost, request: "GET /radio/rest2/savePlayQueue.view?u=user&p=enc:xxxxxxxx&v=2.0.0&c=DSub&id=0000&current=0000&position=0 HTTP/1.1", upstream: "http://[::1]:4040/radio/rest2/savePlayQueue.view?u=user&p=enc:xxxxxxxx&v=2.0.0&c=DSub&id=0000&current=0000&position=0", host: "www.myhostname.tld"

At the time when this error was generated by nginx, I was able to use lynx from that server to connect to localhost:4040/radio, and was served the appropriate content. Even after that the 502 error remains when connecting through nginx.

There is no defined upstream block for this, however the location block is:

location ^~ /radio/ {
        proxy_pass              http://localhost:4040;
        proxy_set_header        Host            $host;
        proxy_set_header        X-Real-IP       $remote_addr;
        proxy_set_header        X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_http_version      1.1;
        proxy_set_header        Connection      "";
#        health_check;       # nginx: [emerg] unknown directive "health_check"
}

I don't want to have to restart nginx every time I boot up. How can I resolve this problem?

chrismeu
  • 133
  • 6
  • Nginx being lightweight, Tomcat just the contrary, could it be that Tomcat isn't ready when nginx checks its availability. If the healthcheck is passive, the backend will be marked as down and stay down. An active healthcheck might help. – Gerard H. Pille Jan 24 '18 at 18:40
  • @GerardH.Pille I believe this is likely the cause of the problem. I had toyed with the idea of modifying systemctl depends options, but this is new to me. I am using vanilla nginx and it does not recognize the 'health_check' directive which seems to be a feature only of plus. Is there an alternative module or method I can use? – chrismeu Jan 24 '18 at 18:50
  • Since it's Tomcat, please also share the `` from your `server.xml`. – Michael Hampton Jan 24 '18 at 19:02

1 Answers1

2

upstream: "http://[::1]:4040/…

Your upstream is probably only listening on IPv4 localhost (127.0.0.1:4040) whereas nginx is trying to connect to IPv6 localhost ([::1]:4040).

lynx works because it tries both.

GUESS: nginx may be failing because it tries both at the start, both fail, then it sticks with IPv6 from there on.

FIX: change upstream to use 127.0.0.1 explicitly or change the upstream to listen on both IPv4 and IPv6.

MikeyB
  • 38,725
  • 10
  • 102
  • 186