Prometheus

Activation

To enable prometheus, simply add the --prometheus flag to the command line arguments of the ingress controller. This will activate the prometheus endpoints in the ingress controller. Its address is the same as the ingress controller and the path is /metrics. The listen port is defined by the option --controller-port which is set to 6060 by default. So ,the expected URL is:

http(s)://<ingress controller ip address or DNS name>:<controller port>/metrics

Security

You can enable a basic authentication access to restrict the access to the prometheus endpoint. To this aim, create a Secret with the pairs of authorized users and their passwords. These passwords must first have been encrypted with a tool like mkpasswd. Example:

$ mkpasswd -m SHA-256

Then point the controller to this secret by adding the prometheus-endpoint-auth-secret data to the controller configmap. Example:

prometheus-endpoint-auth-secret: haproxy-controller/prometheus-credentials

Metrics

On top of the prometheus provided metrics, we added these three ones:

haproxy_reloads_total: The number of haproxy reloads partitioned by result (success/failure)
haproxy_restarts_total: The number of haproxy restarts partitioned by result (success/failure)
haproxy_runtime_socket_connections_total: The number of haproxy runtime socket connections partitioned by object (server/map) and result (success/failure)
haproxy_unable_to_sync_configuration 1 = there's a pending haproxy configuration that is not valid so not applicable, 0 = haproxy configuration applied

Example

Metrics are exposed outside on the haproxy-kubernetes-ingress service, http nodePort, on the /mterics endpoint.

curl <nodeIP>:<haproxy-kubernetes-ingress svc http nodePort>/metrics|grep haproxy_

# HELP haproxy_reloads_total The number of haproxy reloads partitioned by result (success/failure)
# TYPE haproxy_reloads_total counter
haproxy_reloads_total{result="success"} 4
# HELP haproxy_restarts_total The number of haproxy restarts partitioned by result (success/failure)
# TYPE haproxy_restarts_total counter
haproxy_restarts_total{result="success"} 1
# HELP haproxy_runtime_socket_connections_total The number of haproxy runtime socket connections partitioned by object (server/map) and result (success/failure)
# TYPE haproxy_runtime_socket_connections_total counter
haproxy_runtime_socket_connections_total{object="map",result="success"} 4
haproxy_runtime_socket_connections_total{object="server",result="success"} 50
# HELP haproxy_unable_to_sync_configuration 1 = there's a pending haproxy configuration that is not valid so not applicable, 0 = haproxy configuration applied
# TYPE haproxy_unable_to_sync_configuration gauge
haproxy_unable_to_sync_configuration 1