Centralize our servers' logs

Originally created by @intrigeri on #11880 (Redmine)

Right now we have volatile Journal + some persistent log files managed by rsyslog and individual applications. This sometimes makes it painful to debug problems since one has to cross-match info from various sources. It would be nicer if all our logs landed in a single place.

An initial idea to start brainstorming about it would be:

• on each of our systems, send all logs to journald
• configure these journalds to have volatile storage only (that’s the default and what we currently do) and to forward them to a single journald instance running in a central place
• in the central logging location, either have journald store logs in a persistent manner, or forward them to a fancy system like Graylog (that seems much easier to setup than an ELK stack)

Note: For S11, this fits in:

• B.2 - Keep our infrastructure up-to-date and secure: Having centralized logs makes it easier for Sysadmins to have visibility on the current state of systems and services, thus making it easier to conduct diagnoses and spot problems.