[Feature Request] Logging subsystem improvements

[oplehto]

As we scale up our Kapacitor deployment we are incresingly seeing situations where a badly configured task or even corrupt data being sent to a good task can cause a huge flood of logging. The log subsystem should be developed to help us better manage and even avoid these situations.

Some proposals for enhancements:

• [ ] Ability to send logs of each task and topic to separate files
• [ ] Control of log level on a per-task basis
• [ ] Ability to enable log throttling and duplicate log line aggregation with configurable parameters.

[rbetts]

@oplehto If you have specific example log messages that are problematic, we can address those. We are not currently prioritizing a broader enhancement to the logging sub-system.

[oplehto]

The reason for this ticket is exactly that preparing and maintaining such a definitive list of log messages is difficult. The intention would be to be able to contain and manage "unknown unknowns" which could flood our logs and overwhelm our log analytics system.