feat: contextual information about processing for loging using MDC

[csviri]

closes #37

[metacosm]

This looks OK to me but I'm wondering how well it plays with Quarkus.

[csviri]

No idea, next week will take some time and dig deep into quarkus side of things.

[csviri]

Maybe we can merge and see, this how is works its just using ThreadLocals usually in background in logging frameworks. So I guess should not be a big deal.

[csviri]

Maybe we can add this now, in worst case we create a revert commit, if there will be huge issues.

[metacosm]

Another thing is that I feel it's kind of redundant with the monitoring support. Granted, it's probably easier to use than seeing metrics and doesn't quite cover the same use case…

[csviri]

Yes, this is related to log aggregation not to metrics. Like if you have logs in ELK stack and you want to see all logs related to a namespace or a resource. You can easily filter all the related logs.

[csviri]

Actually will investigate how to have better correlation id support in a separate PR. To even better support log aggregation.

[lburgazzoli]

LGTM, however I wonder if we should make such things something a user can opt in

[csviri]

LGTM, however I wonder if we should make such things something a user can opt in

Hmm good question, so if the user should be able to specify the keys and values interested in. Wondering what would be the scope of that, just the custom resource?

Doing the refactor on the other branch, so would rather merge this and create a separate issue and see if there is an elegant way to do it that way.

[lburgazzoli]

LGTM, however I wonder if we should make such things something a user can opt in

Hmm good question, so if the user should be able to specify the keys and values interested in. Wondering what would be the scope of that, just the custom resource?

I meant more a flag to enable/disable MDC entirely (or a dedicated module that enable MDC logging if on the classpath)

[csviri]

@lburgazzoli what would be the benefit of that?
You configure the logging pattern not to log it. SLF4J contains it by default.

[lburgazzoli]

There is usually a cost associated MDC (thread local, maps and other allocations) so if you don't require MDC, then you can turn it off and save some memory

[csviri]

yes, but that part usually doable on the impl level,
for log4j2 at least (on the bottom) https://logging.apache.org/log4j/2.x/manual/thread-context.html

So not sure if it makes that much sense to solve on this level. In case the logger implementation would not handle that sure.

[lburgazzoli]

I don't have any strong opinion, so it is fine with me but as personal preference, I always like more when I have the option to opt in to something that to opt out :)

[csviri]

I don't have any strong opinion, so it is fine with me but as personal preference, I always like more when I have the option to opt in to something that to opt out :)

A completely agree in general. Want to do a round to review for v2 how we are configuring the operator and controller. Will take in mind also this and put it under a feature flag if doable nicely.