HAProxy¶

The haproxy probe monitors HAProxy via its stats CSV HTTP endpoint, collecting session counts, request throughput and error counters per frontend, backend and server component.

Quick start¶

# probes.d/10-haproxy.yaml — each file under probes.d/ is a YAML array of probes
- name: haproxy
  type: haproxy
  params:
    endpoint: http://localhost:8080/stats;csv

Parameters¶

Parameter	Must set	Default	Description
`endpoint`	In practice	`http://localhost:8080/stats;csv`	URL of the stats page in CSV form
`username`	No	-	Basic-auth user when the stats page is protected; empty for none
`password`	No	-	Basic-auth password. A secret: reference it with `${secret:…}`, `${env:…}` or `${file:…}` rather than writing it in the file
`interval`	No	`60`	Seconds between collections
`timeout`	No	`10`	Request timeout in seconds
`instance_name`	No	-	Stable identity of this load balancer; set it when two haproxy probes run on one host

Metrics¶

Metric	Unit	Description
`senhub.haproxy.up`	1	1 when the stats endpoint is reachable and returns valid CSV
`haproxy.sessions.count`	{session}	Current sessions per proxy/component
`haproxy.sessions.total`	{session}	Total sessions since last reload, per proxy/component
`haproxy.bytes.input`	By	Bytes received per proxy/component
`haproxy.bytes.output`	By	Bytes sent per proxy/component
`haproxy.requests.errors`	{error}	HTTP request errors per proxy/component
`haproxy.connections.errors`	{error}	Connection errors per proxy/component

Metrics are tagged with proxy (proxy name) and component (FRONTEND / BACKEND / server name).

Operational notes¶

Enable the stats page in haproxy.cfg: stats enable + stats uri /stats inside a listen stats or frontend block.
Adding stats auth user:password sets the credentials to pass in username/password.
Metric names align with the OpenTelemetry Collector contrib haproxyreceiver where equivalents exist.

Metric reference¶

Every metric this probe can emit. Metric is the OpenTelemetry name the OTLP, Prometheus and Zabbix outputs derive theirs from. Name is what a Nagios check and the API metrics= filter match. PRTG channel is the label PRTG shows, placeholders filled from the series' tags.

Metric	Name	PRTG channel	Unit	Description
`senhub.haproxy.up`	`senhub.haproxy.up`	HAProxy Up	#	1 when the HAProxy stats endpoint is reachable and returns valid CSV
`haproxy.sessions.count`	`haproxy.sessions.count`	HAProxy {proxy} {component} Current Sessions	#	Current number of active sessions (scur)
`haproxy.sessions.total`	`haproxy.sessions.total`	HAProxy {proxy} {component} Total Sessions	#	Total number of sessions since last reset (stot)
`haproxy.bytes.input`	`haproxy.bytes.input`	HAProxy {proxy} {component} Bytes In	B	Total bytes received (bin)
`haproxy.bytes.output`	`haproxy.bytes.output`	HAProxy {proxy} {component} Bytes Out	B	Total bytes sent (bout)
`haproxy.connections.errors`	`haproxy.connections.errors`	HAProxy {proxy} {component} Connection Errors	#	Total connection errors (econ)
`haproxy.requests.errors`	`haproxy.requests.errors`	HAProxy {proxy} {component} Request Errors	#	Total request errors (ereq) — frontends only
`haproxy.responses.errors`	`haproxy.responses.errors`	HAProxy {proxy} {component} Response Errors	#	Total response errors (eresp)
`haproxy.requests.rate`	`haproxy.requests.rate`	HAProxy {proxy} {component} Request Rate	#/s	Current request rate in requests per second (req_rate) — frontends only