Micrometer calls their metrics "Meter primitives", and they support the following:
None of these track time, so the registry is responsible for synthesizing the times at the moment of export.
All micrometer Meter types may contain "tags", which is a collection of key/value pairs.
All of these Tags are converted to New Relic metric Attributes at export time.
Here is how the New Relic registry converts each of these types:
We export the micrometer Gauge as a New Relic Gauge.
Attributes:
- baseUnit: (the gauge's base unit from the id)
- source.type: "gauge"
- description: (the gauge's description)
We export the micrometer TimeGauge as a New Relic Gauge.
Attributes:
- baseUnit: (the gauge's base unit from the id)
- source.type: "time_gauge"
- description: (the gauge's description)
We export the micrometer Counter as a New Relic Count.
- value - delta between current count and previous count (should always be positive)
Attributes:
- baseUnit: (the counter's base unit from the id)
- source.type: "counter"
- description: (the counter's description)
Distribution Summary provides a summary of an event's distribution.
We take a snapshot and marshall it as:
- Snapshot
Count- name: "<name>.count"
- source.type:
distribution_summary
- Snapshot max
Gauge(for maximum event duration)- name: "<name>.max"
- source.type:
distribution_summary
- Snapshot total
Gauge(for total event durations)- name: "<name>.total"
- source.type:
distribution_summary
We also generate one Gauge per snapshot percentile:
- name: "<name>.percentiles"
- source.type:
distribution_summary - percentile: (the actual percentage number)
The framework also generates a Gauge for each histogram bucket.
The FunctionCounter is converted to a New Relic Count object.
Attributes:
- baseUnit: (the function counter's base unit from the id)
- source.type: "function_counter"
- description: (the function counter's description)
There are 3 kinds of timers: Timer, FunctionTimer, and LongTaskTimer.
Aside from the basic Meter interface, these three do not share a common interface.
The micrometer Timer is intended to track a large number of short running events.
We export it as a Count and 3 gauges:
Count:
- name = "<name>.count",
- value = the total number of times stop() has been called on this timer
- (start/end times are tracked between previous export and current export wall time)
Gauge 1:
- name = "<name>.total_time",
- value = the total time of all occurrences
Gauge 2:
- name = "<name>.mean",
- value = the average/mean time of all timer events
Gauge 3:
- name = "<name>.max",
- value = the maximum time of a single event
All 4 metrics have these attributes:
- source.type: "timer"
- baseTimeUnit: (the timer's base time unit)
- description: (the timer's description)
The framework also generates a Gauge for each histogram bucket.
The micrometer FunctionTimer is intended to track two monotonically increasing values: count
(the number of total invocations) and time (total time spent for all invocations).
We export it a Count and 2 gauges:
Count:
- name = "<name>.count",
- value = the total number of times the timer has executed
Gauge 1:
- name = "<name>.total_time",
- value = the total time of all occurrences
Gauge 2:
- name = "<name>.mean",
- value = the average/mean of all events
All 3 metrics have these attributes:
- source.type: "function_timer"
- baseTimeUnit: (the timer's base time unit)
- description: (the timer's description)
The micrometer LongTaskTimer is intended to track the count and total duration of all
in-flight long-running tasks. It is used to time tasks that are still ongoing (because
the plain Timer only reports on completion).
We export it as a two gauges:
Gauge 1:
- name = "<name>.active_tasks",
- value = number of currently active tasks
Gauge 2:
- name = "<name>.total_duration",
- value - The cumulative duration of all active tasks in nanoseconds
Both metrics have these attributes:
- source.type: "long_task_timer"
- baseUnit: (the timer's base unit from the id)
- baseTimeUnit: hard coded to "NANOSECONDS"
- description: (the timer's description)