mcolom
commented
1 year ago The reason why it used $host and not a particular local address was that some requests could come from IPv4, IPv6 and also depend on which particular interface the server was listening to. The requests are resolved locally without any delays by the locally-installed bind9. It's possible that if it's not installed (I haven't checked) it tries to use another one or that it falls into 127.0.0.1 after a while. If setting 127.0.0.1 and removing $host works, I don't have anything to say, it's fine. With a new environment it's possible that it can be done this way without problems. But please check carefully to avoid awful surprises later in production. (more info, updated): I think it was more a problem of nginx than the services. If I recall correctly, what happened without $host and without using bind9 is that you could only reach the services from the outside with the IP, but not with the server's domain. There's something written in the sysadmin doc, in the troubleshooting part. Anyway, let's check it and if it's fine to use 127.0.0.1, I'm fine with that.
this avoids doing DNS requests for nothing