access_log to stdout and error_log to stderr

[oyvindio]

To easily get query logs into the kubernetes fluentd-elasticsearch setup, log access_log to stdout and error_log to stderr.

• Move nginx.conf from the base image into the source tree so that we can modify it.
• When we start nginx, attach the controller's stdout and stderr to the nginx process. For this to work we have to run nginx in the foreground (daemon off). Since we're running nginx in the foreground, use a goroutine to avoid blocking the controller process.
• Set access_log to /dev/stdout and error_log to stderr.

I don't know if you want to merge this as it is a significant change to how logging happens, but we're using this and I think it's useful, so I thought I'd share.

Since we can now edit nginx.conf, I've also added $host, so we can more easily tell which backend we're hitting when using host based Ingress, and $request_time, since that is generally useful, to the main access_log format.

[pleshakov]

Hi,

Thanks for submitting a pull request. I think it's a good idea to make redirecting nginx logs to stdout and stderr by default.

Another way to do it is to insert the following lines in the Dockerfile:

RUN ln -sf /proc/1/fd/1 /var/log/nginx/access.log \
    && ln -sf /proc/1/fd/2 /var/log/nginx/error.log

What we do here is redirecting logs to the stdout and stderr of the process with PID 1, which is the nginx-ingress process. Its stdout and stderr are seen in docker logs. Changes to nginx conf files and ingress controller code are not required in this case.

[oyvindio]

RUN ln -sf /proc/1/fd/1 /var/log/nginx/access.log
&& ln -sf /proc/1/fd/2 /var/log/nginx/error.log

That's a nice solution! 👍

Do you want to merge the access_log format changes, or would you rather keep the defaults? We'll still need nginx.conf in the source tree to do anything with log format I guess

[pleshakov]

Do you already make or need any other customizations of NGINX configuration?

We want to add support for customization of NGINX configuration via Kubernetes resources--ConfigMap and Annotations in an Ingress Resource. So that the user can be able to set proxy_read_timeout, proxy_connect_timeout, etc. Log format is a good candidate for adding the ability to customize. Will it work for you if you configure log format via ConfigMap?

[oyvindio]

We don't have any other customizations besides access_log format right now, but I think we likely will need to do some sort of customization at some point.

ConfigMap would probably be suitable, yeah. That's what the contrib/kubernetes ingress controller is supporting also: https://github.com/kubernetes/contrib/tree/master/ingress/controllers/nginx#custom-nginx-configuration