thresholds in check_graylog_health.py should be configurable

[Gareth Woolridge]

The thresholds for the various components of check_graylog_health.py are hard coded for what is WARBN vs CRITICAL.

For the uncommitted_msgs for instance this is hard coded to 100 - which is rather low for a cluster with any load running within a contended cloud.

Ideally we should expose these values in juju config.

For instance in prod-comms-logging we've cowboyed to 1000 to stop the alert firing in "normal" operation.

[Peter Sabaini (peter-sabaini)]

There's the nagios_uncommitted_{warn|crit} knob in current charm versions, is there anything else that is missing?

[Xav Paice (xavpaice)]

nagios_uncommitted_warn and nagios_uncommitted_crit covers the request in this bug.