Initial Implementation

[thoward]

Overview

This Sensu plugin will perform on-demand monitoring of HAProxy instances. It will collect metrics in Prometheus format, emitting them on STDOUT. Additionally, it may perform analysis to determine the health of those instances, the results of which will be presented as one or more exit status levels.

References

Prior Art:

• Sensu HAProxy Ruby plugin (Ruby, MIT License)
• Telegraf haproxy plugin (Go, MIT License)
• Prometheus HAProxy Exporter (Go, Apache License 2.0)
• Datadog Agent haproxy module (Python, Datadog License)

Misc Docs:

• Datadog integration docs
• Sumo Logic integration docs
• The CSV format HAProxy reports stats with

Configuration

Required

• --url the url to the status endpoint, eg. http://<hostname>:<port>/<status_path> (default: "unix:///run/haproxy/admin.sock")

Optional Backends

• --backends comma-separated list of backends to fetch stats from. If not specified, all HAProxy backends will be included.

Optional Basic Auth Credentials (will be used if specified)

• --admin-user admin username for basic auth
• --admin-pass admin password for basic auth

Optional TLS Config (will be used if tls=custom parameter specified in server uri)

• --tls-ca: Path the the ca.pem file.
• --tls-cert: Path to the cert.pem file.
• --tls-key: Path to the key.pem file.
• --insecure-skip-verify: If true, use TLS but skip chain & host verification (default: false)

NOTE: Datadog has an extensive range of options, which we might want to analyze for beyond-MVP support.

Configuration Logic & Validation

The provided URL must be validated. If supplied, the TLS options must be validated.

Metrics

All metrics should be output in Prometheus format.

MVP to light up the Sumo Logic Dashboards:

Metrics:

• haproxy_active_servers
• haproxy_backup_servers
• haproxy_bin
• haproxy_bout
• haproxy_chkfail
• haproxy_ctime
• haproxy_dreq
• haproxy_dresp
• haproxy_econ
• haproxy_ereq
• haproxy_eresp
• haproxy_http_response_1xx
• haproxy_http_response_2xx
• haproxy_http_response_3xx
• haproxy_http_response_4xx
• haproxy_http_response_5xx
• haproxy_http_response_other
• haproxy_qcur
• haproxy_qmax
• haproxy_qtime
• haproxy_rate
• haproxy_rtime
• haproxy_scur
• haproxy_slim
• haproxy_smax
• haproxy_ttime
• haproxy_weight
• haproxy_wredis
• haproxy_wretr

Tags:

• proxy_cluster
• host
• type: proxy session type. one of frontend, backend, or server
• proxy_system
• component
• proxy
• sv: service name

NOTE: These are the tags/metrics that are seen within the Sumo Logic dashboards, which seem to be based on the Telegraf output. Probably the best place to start looking for details.

For parity w/ Datadog, here's the list of metrics they produce (also available in CSV format).

This metric name map might be useful: https://github.com/DataDog/integrations-core/blob/master/haproxy/datadog_checks/haproxy/metrics.py

Out of scope (for MVP):

• Compatibility and coordination w/ Sardina Systems check
• Implementing legacy check threshold logic. Can be handled by upcoming built-in threshold system.

[echlebek]

Done and dusted.