feat(config): Opt-out of DNS cache

[jjbayer]

Add a config option to disable DNS caching.

DNS caching is "important for K8s infrastructure" (see #2164), but in self-hosted it could have the side effect of caching a negative lookup result when the web container is removed while the relay container keeps running.

ref: getsentry/sentry#109002

[Dav1dde]

Seems like a fine change, but I am not convinced this is going to solve the self hosted problems.

It's very likely that during the first health check Relay has not yet authed and that's why the error message shows up. Maybe we should suppress it, but it doesn't change that the first health check from Kuberentes/docker(-compose) is likely going to fail.

[jjbayer]

I am not convinced this is going to solve the self hosted problems.

It's very likely that during the first health check Relay has not yet authed and that's why the error message shows up.

Initial health check failures are fine, but some users report more persistent problems. Specifically, this is the one that could be explained by a DNS caching problem IMO:

Web Logs: No incoming requests from the relay IP were registered at all.
As I could not find any tool in the relay container to check if it could connect to "http://web:9000/" I added a temporary busybox container to the compose file and tried from there, proving that the internal Docker networking worked (this also showed up with the Docker internal IP of the busybox container as caller_ip in the web logs).

It's worth a try, let's see.

[max-wittig]

Thank you for your efforts here to fix this long standing problem!