Redis logging about servertech-chat

[mzimbres]

The develop branch has logging disabled by default. I also don't know how health checks are supposed to work with XADD since the client can't send pings while it is blocked on XADD etc. You might have to add a second connection to redis that will only does health-checks and cancel other connections in case of a timeout, but this will add some non-trivial complexity. Another option is to use a time interval to XADD that is half of the health checks timeout, so when XADD returns the pings have a chance of being sent before a new XADD is performed.

[anarthal]

As far as I know XADD never blocks - it's XREAD, which we don't use (just because it blocks). So this is just a matter of upgrading Boost.Redis and re-enabling health checks.

I'd still want logged when a redis::connection connects successfully or disconnects from the server, for diagnostics. Is it possible to have these without all the "X bytes written" logs?

[mzimbres]

As far as I know XADD never blocks

Yes, I refered to the wrong command.

• it's XREAD, which we don't use (just because it blocks). So this
  is just a matter of upgrading Boost.Redis and re-enabling health
  checks.

Humm, ok. I thought we would have a connection watching for new elements being added to the Redis stream. If we don't do that, than yes, just re-enable the health checks.

I'd still want logged when a redis::connection connects successfully
or disconnects from the server, for diagnostics. Is it possible to
have these without all the "X bytes written" logs?

Yes, definitely. Boost.Redis default logger has syslog like levels, so we can put connected, resolve etc. messages at a higher level that x-bytes-written. Feel free to open an issue.

I must say that without fmt the logger interface is still messy. It is something I added mostly for myself because as the number of built-in functionality grew e.g. reconnection, health-checks, automatic connect, resolve etc. it became impossible to understand users and my own problems.

Notice that my logger is not thread safe because it uses std::clog to log.