Rename and fix metric names and types

[roidelapluie]

Here are a few changes that should bring this exporter closer to Prometheus best practices.

• phpfpm_scrape_failures is a counter and should end with _total. It has been renamed to phpfpm_scrape_failures_total (https://prometheus.io/docs/instrumenting/writing_exporters/#naming)
• phpfpm_start_since has been renamed to phpfpm_start_time_seconds to match standard process_start_time_seconds and contain unit. It has been changed to a gauge (https://www.robustperception.io/are-increasing-timestamps-counters-or-gauges).
• phpfpm_accepted_connections is a counter and should end with _total. It has been renamed to phpfpm_accepted_connections_total.
• phpfpm_idle_processes, phpfpm_active_processes have been changed to phpfpm_processes{state="active"} and phpfpm_processes{state="idle"}.
• phpfpm_total_processes has been dropped because it can be found as sum without(state) phpfpm_processes.
• phpfpm_max_children_reached is a counter and should end with _total, so it has been renamed to phpfpm_max_children_reached_total.
• process_last_request_memory should contain the unit, so it has been renamed to process_last_request_memory_bytes.
• process_request_duration has been normalized to seconds and renamed phpfpm_process_last_request_duration_seconds.

[itcsoft54]

Rename metric create breaking change with existing use of phpfpm_exporter. Need to deprecate old metric instead of remove (or rename) it

[roidelapluie]

Certainly. The main problem is process_request_duration which now uses the normalized second (https://prometheus.io/docs/practices/naming/#base-units).

For the rest, we could offer recording rules or relabel configs to ease the transition and release a 3.0 maybe.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[gerardnico]

All max metrics such as phpfpm_max_children_reached should be a gauge, no? I don't understand why they are set as counter. They are not going up for every request and they can go down.

[gerardnico]

Note that on remote_write, the type of metrics metadata seems to be not send, you can rename them:

- url: https://metric-api.newrelic.com/prometheus/v1/write?X-License-Key=...
  write_relabel_configs:
  - source_labels: [__name__]
    regex: ^my_counter$
    target_label: newrelic_metric_type
    replacement: "counter"
    action: replace

Ref