Prometheus

PreviewProxy can expose a Prometheus-compatible metrics endpoint on a dedicated address, separate from the main proxy port.

Enabling the metrics endpoint

Set PP_PROMETHEUS_BIND to the address and port to listen on:

PP_PROMETHEUS_BIND=:9464

Or use the CLI flag:

previewproxy --prometheus-bind :9464

The endpoint is available at GET /metrics on the configured address. If PP_PROMETHEUS_BIND is not set, the metrics endpoint is disabled.

By default, metrics have no prefix. Set PP_PROMETHEUS_NAMESPACE to add a prefix to all metric names:

PP_PROMETHEUS_NAMESPACE=previewproxy

With this set, requests_total becomes previewproxy_requests_total, and so on for all metrics.

Metric	Labels	Description
`requests_total`	-	Total HTTP requests processed
`status_codes_total`	`status`	Response count by HTTP status code
`errors_total`	`type`	Errors by type (e.g. `timeout`, `upstream`)
`cache_hits_total`	`layer`	Cache hits by layer (`memory`, `disk`)
`cache_misses_total`	`layer`	Cache misses by layer (`memory`, `disk`)

Metric	Labels	Description
`request_duration_seconds`	-	Full request latency in seconds
`request_span_duration_seconds`	`span`	Request latency broken down by span
`buffer_size_bytes`	-	Download buffer sizes in bytes

Metric	Labels	Description
`workers`	-	Configured max concurrent requests
`requests_in_progress`	-	Requests currently being handled
`images_in_progress`	-	Images currently being processed
`workers_utilization`	-	Worker utilization percentage (`requests_in_progress / workers * 100`)
`buffer_default_size_bytes`	-	Calibrated default buffer size in bytes
`buffer_max_size_bytes`	-	Calibrated maximum buffer size in bytes
`cache_memory_size_bytes`	-	Current memory cache size in bytes
`cache_disk_size_bytes`	-	Current disk cache size in bytes
`cache_entries`	`layer`	Current cache entry count by layer

scrape_configs:
  - job_name: previewproxy
    static_configs:
      - targets: ["localhost:9464"]