ptemplate-observability.md

Monitoring and Metrics

Introduction

Kubernetes is designed to talk use Prometheus to gather metrics on what is going on. These metrics can then be passed to Grafana for logging and graphing. This means there is good support for gathering and communicating values in the golang libraries.

There are two things that need to be done in the operator, firstly registering a data structure for each metric to be gathered, and secondly actually gathering the metrics. The actual sending of the metrics to Prometheus is then handled automatically.

Metrics in the Ptemplate Operator

Registering Metrics

In the Ptemplate operator we use a separate golang package to register the metrics we are going to be gathering. This simply creates a new object of the appropriate type for each metric, passing the name of the metric and some help text. Then it calls the MustRegister() function for each of them.

PtemplateGauge = prometheus.NewGauge(
    prometheus.GaugeOpts{
        Name: "Ptemplate_consumers",
        Help: "Number of existing Ptemplate consumers",
    },
)

...

metrics.Registry.MustRegister(PtemplateGauge)

And we then initialise this package in the init() function of main.go

func init() {
...
monitoring.RegisterMetrics()
}

Reporting Metrics

To then gather the metrics we use that Prometheus object anywhere else in the code we need to publish the metric

monitoring.PtemplateGauge.Set(float64(len(podNames)))

Links

Monitoring and Observability

Prometheus golang client documentation

OpenShift Documentation: Managing metrics