Add blockdiff and blocknumber gauges for Prometheus

[joaquincasares]

Description

Implement Prometheus Gauges and add blockdiff and blocknumber metrics for the exporter using a collector function. Collector functions allow us to leave business logic next to related business logic within the same file, while exposing a group of gauges that should only be inspected at export/scrape time.

Tests

How will this change be monitored? Are there sufficient logs?

[dmanjunath]

@joaquincasares just so i understand correctly what's happening here is

1. Create a new type of collector for gauges
2. Create and register gauges to run on sweep
3. When the sweep happens, run get_health and return 1 if unhealthy, 0 otherwise?

[csjiang]

Looks cool! I am still gathering context here but added a comment/q.

[csjiang]

Maybe we abstract these out into addPrometheusMetric('gauges'), addPrometheusMetric('histograms')?

[joaquincasares]

Phew! Thanks! I knew I didn't like the gauge=false bit since I would only extend to two types, but figured to punt that design choice until later.

I'll implement your suggestion, thanks!

[joaquincasares]

Heh, I misread your suggestion, but hopefully the latest change is what you had in mind?

[joaquincasares]

@joaquincasares just so i understand correctly what's happening here is

1. Create a new type of collector for gauges
2. Create and register gauges to run on sweep
3. When the sweep happens, run get_health and return 1 if unhealthy, 0 otherwise?

1. Correct.
2. We don't have to collect gauges at exporter time, but I figured it would be cleaner to create our own "health check exporter" instead of having PrometheusMetrics objects within get_health(). Since health checks are fired routinely a collector isn't required here, but does make the code cleaner. For other gauges, however, a collector may be required.
3. Whoops! Fixed the double negative logic. Thanks!