Task time metric buckets entirely reside in +Inf

[Lemmmy]

The computercraft_server_tasks_s_bucket and computercraft_task_time_s_bucket metrics only ever have data in the +Inf bucket. My suspicion is that the metrics system collects data in nanoseconds, but since the buckets are defined in seconds, all the data always ends up being higher than 0.25. This results in the metrics being unusable for histogram and heatmap visualisations.

For example:

computercraft_server_tasks_s_bucket{computer_id="1439",le="5.0E-4",} 0.0
computercraft_server_tasks_s_bucket{computer_id="1439",le="0.001",} 0.0
computercraft_server_tasks_s_bucket{computer_id="1439",le="0.005",} 0.0
computercraft_server_tasks_s_bucket{computer_id="1439",le="0.01",} 0.0
computercraft_server_tasks_s_bucket{computer_id="1439",le="0.025",} 0.0
computercraft_server_tasks_s_bucket{computer_id="1439",le="0.05",} 0.0
computercraft_server_tasks_s_bucket{computer_id="1439",le="0.1",} 0.0
computercraft_server_tasks_s_bucket{computer_id="1439",le="0.25",} 0.0
computercraft_server_tasks_s_bucket{computer_id="1439",le="+Inf",} 108012.0

cc-prometheus/common/src/main/java/cc/tweaked/prometheus/collectors/ComputerFieldCollector.java

Lines 43 to 46 in 109de72

	serverTime = buildField(Metrics.SERVER_TASKS, Histogram.build())
	.unit("s")
	.buckets(0.0005, 0.001, 0.005, 0.01, 0.025, 0.05, 0.10, 0.25)
	.register(registry);

https://github.com/cc-tweaked/CC-Tweaked/blob/54b7366b2ddbb0ee024e0b06ff3b9d300580e554/projects/core/src/main/java/dan200/computercraft/core/metrics/Metrics.java#L14-L15

This can be tested on the sc3 staff-analytics with queries such as sum by (le) (computercraft_task_time_s_bucket[1m]) and sum by (le) (rate(computercraft_server_tasks_s_bucket[1m])).

This does not affect CC's own commands as the times are always formatted back as seconds:

https://github.com/cc-tweaked/CC-Tweaked/blob/54b7366b2ddbb0ee024e0b06ff3b9d300580e554/projects/core/src/main/java/dan200/computercraft/core/metrics/Metric.java#L108-L110

[Lemmmy]

❤️