master node syslog server custom port/proto issue

[Bogdan Dobrelya]

- By default, we have rsyslog server at the master node configured at the bootstrap_admin_node stage only, and there is no way to provide any additional ports to listen but default one (514). Note: it is easy to make it to listen both tcp and udp though.
- This is an issue, because there is an option in UI to specify additional rsyslog server, proto and port to listen (as well as to send to for its clients) but it would work only if other (i.e. not the master node itself) rsyslog server was chosen and if it listens the port and proto specified in UI. In case of the user (or CI tests, for example) had specified custom proto and(or) port for the same master node - the server would not be reconfigured to listen for it, thus clients will send their logs to wrong endpoint.

Possible solutions:
1) to restrict the user (CI tests) from using the same master node's IP/localhost as additional rsyslog server in UI.
2) to make master node's rsyslog server reconfigurable as well (will require to add multiple listen ports support for rsyslog::server puppet class)

[Bogdan Dobrelya (bogdando)]

https://review.openstack.org/70966

[Bogdan Dobrelya (bogdando)]

The patch suggested in 70966 is about to implement the option #2
TODO make rsyslog server at the master node reconfigurable by syslog UI settings updates

[Dmitry Pyzhov (dpyzhov)]

I see no issue here. If user wants to add more rsyslog destinations, he needs to configure remote servers by hand. And he can create additional server on master node, we should not forbid it. For example, he can set up forwarding server or some analytics tool. We should not care about destination health.