Reporting metrics

Paddler provides reporting metrics to help you monitor your setup. You can either scrap the /metrics endpoint to obtain metrics in the Prometheus format, or ask Paddler to push StatsD metrics to a given address.

Provided metrics

You can obtain the following metrics:

• paddler_slots_processing - the number of slots currently processing requests
• paddler_slots_total - number of total available slots
• paddler_requests_buffered - the number of requests currently buffered

Metrics prefix

By default, Paddler reports the metrics with the paddler_ prefix, but you can change this prefix when running the balancer with the --statsd-prefix flag.

For example, if you set the prefix to foo_ running this command:

paddler balancer --inference-addr 127.0.0.1:8061 --management-addr 127.0.0.1:8060 --statsd-prefix foo_

you will receive the following:

• foo_slots_processing
• foo_slots_total
• foo_requests_buffered

Prometheus format metrics

Paddler exposes the /metrics endpoint you can scrap to obtain metrics in the Prometheus format.

Method: GET
Path: /metrics

StatsD metrics

If enabled, Paddler will send metrics to your StatsD server as gauge values. You can enable this when running the balancer with the --statsd-addr flag, for example:

paddler balancer --inference-addr 127.0.0.1:8061 --management-addr 127.0.0.1:8060 --statsd-addr 127.0.0.1:8125