Startup with wrong configuration

[Rillke]

Hi,

today, after a VM reboot, this nginx-proxy did startup with an inappropriate configuration leading to requests being directed to the wrong container. After manually running docker-compose restart nginx-proxy the correct configuration has been generated.

Here is such a request (log):

nginx-proxy_1  | nginx.1     | 2021/12/20 06:37:01 [error] 46#46: *21639 connect() failed (111: Connection refused) while connecting to upstream, client: 1.1.1.1, server: a.example.com, request: "GET /search/episode.xml?limit=6&offset=0 HTTP/1.1", upstream: "http://172.19.0.6:8080/search/episode.xml?limit=6&offset=0", host: "a.example.com"

docker network inspect:

            "3dc64bbc858baa0412cb68xxx": {
                "Name": "software-docker_container1_1",
                "EndpointID": "016d71c705xxx",
                "MacAddress": "02:xx:xx:13:00:06",
                "IPv4Address": "172.19.0.6/16",
                "IPv6Address": ""
            },
            "a0ee561716141553677c73xxx": {
                "Name": "software-docker_container2_1",
                "EndpointID": "cea7d59828675xxx",
                "MacAddress": "02:xx:xx:13:00:02",
                "IPv4Address": "172.19.0.2/16",
                "IPv6Address": ""
            },
            "bed9bdc7e0398fdxxx": {
                "Name": "software-docker_nginx-proxy_1",
                "EndpointID": "c5bb23758afxxx",
                "MacAddress": "02:xx:xx:13:00:03",
                "IPv4Address": "172.19.0.3/16",
                "IPv6Address": ""
            },

While the request for a.example.com was supposed to go to 172.19.0.2:8080, it was sent to 172.19.0.6:8080 (this is what I get from the above request log's upstream field).

Here are the relevant variables for each container:

software-docker_container1_1:

VIRTUAL_PORT=9000
VIRTUAL_PROTO=fastcgi
VIRTUAL_HOST=3.3.3.3

(container1's VIRTUAL_HOST equals the host's public IP address)

software-docker_container2_1:

VIRTUAL_PORT=8080
VIRTUAL_HOST=a.example.com

software-docker_nginx-proxy_1:

DEFAULT_HOST=a.example.com

That's the first time I'm experiencing this issue in production. Have there been any regressions recently or am I using it wrong?

Thanks!

[buchdag]

Hi

(container1's VIRTUAL_HOST equals the host's public IP address)

I don't recommend doing this. Even if it does work, it's never been officially supported nor tested, and is not the intended use of the container.

That said I've experienced issues similar to yours (a request being proxied to an correct container, with the misconfiguration being after a container restart) in the past without using IP in place of proper host name in VIRTUAL_HOST. I've never been able to reliably reproduce the issue and it's always been a very uncommon occurrence.

To my knowledge there hasn't been any recent known regression that could have caused this.