OpenMetrics

StackState for Kubernetes troubleshooting

Overview

StackState Agent V2 can be configured to retrieve metrics from an OpenMetrics endpoint and push these to StackState.

Setup

Installation

The OpenMetrics check is included in the [Agent V2 StackPack].

Configuration

To enable the OpenMetrics integration and begin collecting metrics data from an OpenMetrics endpoint, the OpenMetrics check must be configured on StackState Agent V2. The check configuration provides all details required for the Agent to connect to your OpenMetrics endpoint and retrieve the available metrics.

Kubernetes, OpenShift

1. Deploy the Agent on your Kubernetes or OpenShift cluster.

2. Add the annotations below when launching a pod that exposes metrics via an OpenMetrics endpoint. Add the following:

   • <CONTAINER_NAME> - the name of the container that exposes the OpenMetrics. It's possible to process multiple endpoints in a single pod (that's why there is a list in the JSON).

   • prometheus_url - the path (often just metrics) and port at which the OpenMetrics endpoint is exposed.

   • namespace - all metrics collected here will get this as a dot-separated prefix.

   • metrics - use ["*"] to collect all available metrics. It's also possible to specify a list of metrics to be fetched. This should either be a string representing the metric name or a mapping to rename the metric<EXPOSED_METRIC>:<SENT_METRIC>

     ...
     metadata:
       annotations:
         ad.stackstate.com/<CONTAINER_NAME>.check_names: '["openmetrics"]'
         ad.stackstate.com/<CONTAINER_NAME>.init_configs: '[{}]'
         ad.stackstate.com/<CONTAINER_NAME>.instances: |
           [ 
             {
               "prometheus_url": "http://%%host%%:<METRICS_PORT>/<METRICS_PATH>",
               "namespace": "<METRICS_NAMESPACE>", 
               "metrics": ["*"] 
             } 
           ]
     ...
     # This already exists in the pod spec, the container name needs to match the container that is exposing the openmetrics endpoint
     spec:
       containers:
        - name: <CONTAINER_NAME>
     ...

3. You can also add optional configuration and filters:

   • prometheus_metrics_prefix - prefix to add to exposed OpenMetrics metrics.

   • health_service_check - send a service check <NAMESPACE>.prometheus.health reporting the health of the OpenMetrics endpoint. Default true.

   • label_to_hostname - override the hostname with the value of one label.

   • label_joins - target a metric and retrieve it's label via a 1:1 mapping

   • labels_mapper - rename labels. Format is <LABEL_TO_RENAME>: <NEW_LABEL_NAME>.

   • type_overrides - override a type in the OpenMetrics the payload or type an untyped metric (these would be ignored by default). Supported <METRIC_TYPE> are gauge, count and rate. Format is <METRIC_NAME>: <METRIC_TYPE>.

   • tags - list of tags to attach to every metric, event and service check emitted by this integration.

   • send_histograms_buckets - send the histograms bucket. Default true.

   • send_monotonic_counter - set to true to convert counters to a rate (note that two runs are required to produce the first result). Set to false to send counters as a monotonic counter. Default true.

   • exclude_labels - list of labels to be excluded.

   • prometheus_timeout - set a timeout for the OpenMetrics query.

   • ssl_cert - If your OpenMetrics endpoint is secured, enter the path to the certificate and specify the private key in the ssl_private_key parameter, or give the path to a file containing both the certificate and the private key.

   • ssl_private_key - required if the certificate linked in ssl_cert doesn't include the private key. Note that the private key to your local certificate must be unencrypted.

   • ssl_ca_cert - the path to the trusted CA used for generating custom certificates.

   • extra_headers - a list of additional HTTP headers to send in queries to the OpenMetrics endpoint. Can be combined with autodiscovery template variables. For example, "Authorization: Bearer %%env_TOKEN%%".

4. Wait for the Agent to collect data from the OpenMetrics endpoint and send it to StackState.

Data collected

Metrics

By default, all metrics are retrieved from the specified OpenMetrics endpoint. To optimize performance, a maximum of 2000 metrics will be retrieved. If the check is attempting to retrieve more than 2000 metrics, add a metrics filter to the configuration to ensure that all important metrics can be retrieved within the limit.

Retrieved metrics won't automatically be mapped to topology elements but they can be browsed using the telemetry inspector and eventually added to components via a metric binding.

Events

The OpenMetrics integration doesn't retrieve any events data.

Traces

The OpenMetrics integration doesn't retrieve trace data.