madwire / dockercloud-nginx-proxy Goto Github PK

Hi,
this proxy works just fine for most of my applications. Now we also want to deploy applications like atlassian jira within this proxy, where we don't have the possibility to set the internal ports, e.g. confluence runs on port 8090.

It would be very useful to have a VIRTUAL_PORT environment variable as well.

It's reconsigned the number of nodes but doesn't create the same number of containers.

Hello,

I couldn't find a proper example of using FORCE_SSL option with a SSL I gathered from LetsEncrypt?

How should I do it?

Best

This looks like a great project. Do you have a succinct list of pros and cons vs docker clouds standard recommendation of dockercloud/haproxy?

Hi!

I have been trying to get this nginx-proxy up and running for a couple of days now, and I am not certain what I am missing.

The setup consists two services currently on the same node.

When I redeploy my other service I can see from the nginx service logs that it receives the events and reinitialises and rebuilds the nginx configuration. But from what I can see in the logs it only recognizes the nginx service itself and does not build a configuration for service that actually just redeployed.

Mind you that I have tried this setup on different nodes and on the same node and the result is the same. I did of course specify the VIRTUAL_HOST environment variable on the service i wish to proxy to, and my other service is listening on port 80, but the port is not published.

So currently I gathered one of two possibilities for this not working.

1. Networking
2. Bug due to api changes in docker cloud.

Hope you could point me in the right direction or give me some genereal direction to where I should start debugging.

Dump of the nginx service logs:

dockercloud.1 | I, [2018-03-09T12:35:26.255388 #11] INFO -- : Init Nginx config
dockercloud.1 | I, [2018-03-09T12:35:29.653237 #11] INFO -- : dockercloud-nginx-proxy-ab65de63: ["10.7.0.3"]
dockercloud.1 | I, [2018-03-09T12:35:29.656725 #11] INFO -- : Connecting to wss://ws.cloud.docker.com/api/audit/v1/events
dockercloud.1 | 2018/03/09 12:35:29 [notice] 25#25: signal process started
dockercloud.1 | I, [2018-03-09T12:35:30.081335 #11] INFO -- : Connected!
dockercloud.1 | I, [2018-03-09T12:35:55.174832 #11] INFO -- : Service: - is Redeploying...
dockercloud.1 | I, [2018-03-09T12:36:12.257614 #11] INFO -- : Service: - is Running!
dockercloud.1 | I, [2018-03-09T12:36:20.890809 #11] INFO -- : dockercloud-nginx-proxy-ab65de63: ["10.7.0.3"]
dockercloud.1 | I, [2018-03-09T12:36:20.891694 #11] INFO -- : Writing new nginx config
dockercloud.1 | I, [2018-03-09T12:36:23.895399 #11] INFO -- : Reloding Nginx...
dockercloud.1 | 2018/03/09 12:36:23 [notice] 34#34: signal process started

I forked your repo and added a folder with a Dockerfile, along with your cloned ruby files, for OpenResty 1.9.7.3, located in openresty folder.

I ran your version without any trouble in Docker Cloud. However, when running version with OpenResty (conceivably not much should change) I'm having trouble obtaining the proper configuration settings for ../nginx/nginx.conf and ../nginx/conf.d/default.conf, seems like the default configuration file is not there and does not have the include statement. Fortunately, the log file outputs messages where I can see that the upstream servers are resolving correctly. So no issues with nginx.conf.erb and dockercloud.rb, just not sure how you got the first part of /etc/nginx/nginx.conf customized with the include statement since the FROM directive in your Dockerfile seems to pull from the standard Nginx 1.9.0 image.

For OpenResty image, I based Dockerfile in large part from ficusio/openresty, which I also ran without any issues in Docker Cloud (although it doesn't provide service discovery out of the box).

Any ideas on how this could be configured?

Once it's fixed I'll send you a PR.

In docker stacks you can name your services "web" or "database". The absolute service name will then be e.g. "my-cool-stack.web". If I deploy two of those, the service names are individual for docker itself, not for the dockercloud-nginx-proxy. I currently have a "web" for every stack I booted which leads to that just the first one works.

Since it's just an internal marker, can we use service.uuid instead of service.name? (or maybe service.name + "-" + service.uuid?)