Http check specific URL - documentation

[yaoweibin]

Http check specific URL - documentation

[vaijab]

I believe you're after the following parameter for specifying custom URL:

check_http_send "GET /server-health.asp HTTP/1.0\r\n\r\n";

[villanus]

@vaijab I just tried what you advised and it did not work.

[villanus]

I am trying to follow this advice, but have no idea what I am doing:
http://stackoverflow.com/questions/6686261/what-at-the-bare-minimum-is-required-for-an-http-request

[villanus]

I am having a really hard time from this. I have 3 iis backend servers running on azure. When I turn one of the servers off, this module fails to recognize that the server is off. The reason for this is Azure actually keeps what they call a "cloud service" open and the page still loads. If i completely stop the service then the module assumes my ip address is 0.0.0.0 and again the check comes back healthy.

The original CEP21 version of this module had some features which allowed you to send custom get reuqest and even had healthcheck_expected. I dont really care if expected does not work, but how can i simply send a get request so that if the page doesnt load then it will fail?

How can healtcheck can just check status.php or status.asp and if that gives a status of 304 or 200 then we are ok?

How can i make sure when upstream servers are entered as dnsnames (ww1.mydomain.com, ww2.mydomain.com), when they are down healthcheck doesnt associate their ip address with 0.0.0.0. To test this, just create an upstream server by hostname, and a cname to a server that doesnt exist.

[villanus]

Please also see this
https://github.com/cep21/healthcheck_nginx_upstreams/blob/master/sample_ngx_config.conf

[yaoweibin]

Have you specified the virtual host in your IIS backend servers? If so, you should add the host header like this:

check_http_send "GET /server-health.asp HTTP/1.0\r\nHost: yourname.com\r\n\r\n";

The 2xx or 3xx response will be marked as healthy by default.

[villanus]

Thanks! That worked brialliantly. One thing that would be nice is if rather than putting the hostname "yourname.com" if it would just pull the hostname from the upstream servers. Something like this:
check_http_send "GET /server-health.asp HTTP/1.0\r\nHost: upstream\r\n\r\n";

That way say if each upstream is configured with a distinct virtual host (ww1, ww2, ww3), it would just poll the appropriate url during the health check and result in this:
check_http_send "GET /server-health.asp HTTP/1.0\r\nHost: ww1.mybackenedservers.com\r\n\r\n";
check_http_send "GET /server-health.asp HTTP/1.0\r\nHost: ww2.mybackenedservers.com\r\n\r\n";

[yaoweibin]

I'm afraid it‘s difficult to know if the user configures the right virtual host name for this upstream. It should be flexiable to add the host in the check_http_send directive.