Prometheus Instrumentation Example in Go

This project implements a simple web server/client. The webserver code includes:

• instrumentation of Prometheus client
• configuration of custom metrics
• simulation of status code, duration, and payload size for each HTTP request

The client runs an endless loop of sending HTTP requests to the web server.

Build

make

Queryring

When deployed with Promethus server, the following PromQL queries might be helpful to get webserver operational status:

Average process memory use:

avg_over_time(go_memstats_alloc_bytes[1m])

Request rate:

sum(rate(http_requests_total[1m]))

Error rate:

sum(rate(http_requests_total{status_code!~"2.."}[1m]))

Error ratio (in %):

sum(rate(http_requests_total{status_code!~"2.."}[1m])) / sum(rate(http_requests_total[1m])) * 100

Average request duration (grouped by status code):

sum(rate(request_duration_seconds_sum[1m])) by (status_code) / sum(rate(request_duration_seconds_count[1m])) by (status_code)

85-percentile of request duration (in sec.):

histogram_quantile(0.85, sum(rate(request_duration_seconds_bucket[1m]))by(le))

99-percentile of successful request duration (in sec.):

histogram_quantile(0.99, sum(rate(request_duration_seconds_bucket{status_code="200"}[1m]))by(le))

Average response payload size (in bytes):

sum(rate(response_size_bytes_sum[1m])) / sum(rate(response_size_bytes[1m]))

90-percentile of average response payload size (in bytes) for successful requests:

avg_over_time(response_size_bytes{quantile="0.9", status_code="200"}[1m])