write_prometheus fails to bind when the host has no non-loopback IP addresses about collectd HOT 2 CLOSED

Hi,

thank you for the detailed writeup, this is excellent!

I wonder if you are running into an ordering issue. I have seen this with other daemons too; the resolution would be to launch collectd when all networks are up. The underlying issue is in my understanding, that there is no (good?) way to update the addresses to listen on when collectd (or other processes) are already running.

from collectd.

The underlying issue is in my understanding, that there is no (good?) way to update the addresses to listen on when collectd (or other processes) are already running.

There is, it's called binding to the wildcard address (0.0.0.0, ::, INADDR_ANY, etc.).

If collectd just bound to hardcoded 0.0.0.0 before all networks are up, all would be fine.
However, it tries to obtain the 0.0.0.0 sockaddr through getaddrinfo instead.
And getaddrinfo would resolve it just fine, even before all networks are up, if it wasn't for AI_ADDRCONFIG

resolution would be to launch collectd when all networks are up

Other daemos that bind to 0.0.0.0 or 127.0.0.1 work just fine when they launch before all networks are up.

If I was trying to bind my LAN address (eg. 192.168.1.20) before all networks are up, then that would be an ordering issue. the wildcard address should work just fine regardless of networks being up.

from collectd.