On podman containers, /etc/resolv.conf file is generated based on the host configuration. In my system, I must have some IPv6 config enabled, and that leads podman to generate this /etc/resolv.conf in the apicast container:

nameserver 10.0.2.3
nameserver 192.168.100.1
nameserver fe80::1%2

Note that the last line is a link-local IPv6 address in zone-id format (learn more). Other examples of such addresses are fe80::1%eth0 or fe80::1%enp0s20f0u1.

Trying to run an apicast container which such type of IPs in the resolver fails with the error:

nginx: [emerg] invalid IPv6 address in resolver "[fe80::1%2]:53" in /tmp/lua_l9C4j9:52

In Fedora 37, this can be reproduced by running:

podman run --name apicast --rm -p 8080:8080 --add-host=host.docker.internal:10.0.2.2 --net slirp4netns:allow_host_loopback=true -e THREESCALE_PORTAL_ENDPOINT="http://apicastsecret@10.0.2.2:3008/master/api/proxy/configs" -e THREESCALE_DEPLOYMENT_ENV=staging -e BACKEND_ENDPOINT_OVERRIDE="http://10.0.2.2:3001" quay.io/3scale/apicast:latest

I'm not sure of this, but I suspect that some regular expressions in the apicast code could be the cause. A regular expression like the next one could be more appropriate:

[a-f\d:]+(%.+)?

I'm having to disable IPv6 via podman parameter in order to get apicast working in my machine.

I'm not sure how to test this in my machine, so, could anybody take a look?

UPDATE
I attached the content of the lua_l9C4j9 file, which seems to be a NGINX configuration file, which line 52 looks like this:

resolver 10.0.2.3:53 192.168.100.1:53 [fe80::1%2]:53;

Is this file generated by our code? it seems the format is incorrect