Missing port in address when using lookup_srv

[jamesmogg]

1. Caddy version (V2.1):

2. How I run Caddy:

I'm running caddy in Docker on alpine from a simple run command, im configuring it using the API from a nodejs app, using axios.

a. System environment:

Docker Alpine 3.1

b. Command:

docker run -d -p 8080:80 -p 9999:9999 --network my-net

c. Service/unit/compose file:

paste full file contents here

d. My complete Caddyfile or JSON config:

const sta = {
  admin: {
    listen: "0.0.0.0:9999",
  },
  logging: {},
  apps: {
    http: {
      servers: {
        example: {
          listen: [":80"],
          routes: [
            {
              match: [{ path: ["/main"] }],
              handle: [
                {
                  body: "Hello, world!",
                  handler: "static_response",
                },
              ],
            },
            {
              match: [{ path: ["/other"] }],
              handle: [
                {
                  handler: "reverse_proxy",
                  transport: {
                    protocol: "http",
                    tls: {
                      insecure_skip_verify: true,
                      server_name: "node-express1",
                    },
                  },
                  upstreams: [
                    { lookup_srv: "http://node-express1:8522" },
                  ],
                },
              ],
            },
          ],
        },
        logs: {},
      },
    },
  },
};

3. The problem I’m having:

In the documentation it says, “If DNS SRV records are used for service discovery with this upstream, specify the DNS name for which to look up SRV records here, instead of specifying a dial address.”

I want to use the docker name of my container , in this case node-express1, but if i put that in the looup_svr variable it give an error saying the port is missing, it seems if i don’t include the dial variable it will not take the config and gives the port missing error. If i put anything in dial it’ll take the config but wont work. I also cant get logging to work, i followed the forum suggestion and added logs:{}, but it doesnt return anything when i curl it.

I can get a response from the node-express1 container via hostname from the caddy container command line.

4. Error messages and/or full log output:

loading config: loading new config: loading http app module: provision http: server example: setting up route handlers: route 1: loading handler modules: position 0: loading module 'reverse_proxy': provision http.handlers.reverse_proxy: missing port in address

5. What I already tried:

I have tried adding the transport section i found someone else suggesting was the problem, that didn’t help, I’ve tried putting dial:“:8852” in to give it a port, that will allow the config to load but it doesn’t work. I cant really think what else to try and without logging its hard to know what to do.

Incidentally i know i should’nt(and wont) be exposing my admin port, but its easier to upload configs from my machine to docker with it open and i can keep track of the million confings ive tried.

this is the command i used to try and get the logs

 curl localhost:9999/config/apps/http/servers/example/logs/ -X POST -H "Content-Type: application/json" -d '{}'

it just returns nothing

6. Links to relevant resources:

Tried this

reverse_proxy: Add support for SRV backends

tried this to get loggin working, didnt work

V2: how do I get access logs?

Thanks for putting this self-contained and working snippet, it helped me a lot! I had trouble finding how to have access logs with Caddy2 (coming from Caddy 1). I found this other post Logging doesn't work in 2.0 beta 13? - #2 by francislavoie which hints at using the API like matt suggested here, but after trying multiple times I didn’t found the right way to do it. Providing a working snippet is great!

[francislavoie]

Unfortuately it’s a known issue, but there’s a fix in the works:

https://github.com/caddyserver/caddy/pull/3756

[jamesmogg]

Thanks for your reply, Sorry I didn’t find the bug post myself. I tend to assume it’s my mistake rather than a bug.