Removing possibly outdated code from TrellServer.cs

[doug-jacob]

Following a conversation with Matthew during recent Trell cleanup, there was one notable section of commented-out code which we aren't sure is still relevant, so its removal is its own PR

EDIT: the purpose of the entire method was in question, so this PR changed to evaluate it wholistically

[isaksky]

I think we'd need to do the work first to see if that is needed. E.g., how are workers logging now - is that ok, or should it be centralized?

[oconnor0]

If it does make sense to remove Log, it will make sense to remove LogRequest and any other messages/types it depends on.

[oconnor0]

Structurally, each worker process loads the implementation of ITrellLogger defined in the config and logs via that. This direct logging makes sense, conceptually, to me, as it allows the plugin to define how logging is done - stdout, file, Redis, etc. but I don't know if it's working OK in the places it's being used.

[isaksky]

This direct logging makes sense, conceptually, to me, as it allows the plugin to define how logging is done

Well the plugin could be loaded anywhere, so that doesn't seem like a difference in capability to me.

One downside of the way we are doing it now is that the plugin may use scarce resources, such like connections, so it doesn't scale as well. E.g., a setup running 50 workers now needs 50 connections/connection-pools to redis or postgres (for example), when maybe 1 would do.

Also the overhead of constantly starting/stopping those.

[oconnor0]

We could define a LogViaServer : ITrellLogger implementation that uses the rpc Log on the server to log from the worker and then provide a way to specify whether the worker should load that logger or the same one as the server.

Though maybe that's unnecessary complexity/work at this point.