Monitor Astra Trident

07/26/2024 Contributors

Astra Trident provides a set of Prometheus metrics endpoints that you can use to monitor Astra Trident’s performance.

The metrics provided by Astra Trident enable you to do the following:

Keep tabs on Astra Trident's health and configuration. You can examine how successful operations are and if it can communicate with the backends as expected.
Examine backend usage information and understand how many volumes are provisioned on a backend and the amount of space consumed, and so on.
Maintain a mapping of the amount of volumes provisioned on available backends.
Track performance. You can take a look at how long it takes for Astra Trident to communicate to backends and perform operations.

By default, Trident's metrics are exposed on the target port 8001 at the /metrics endpoint. These metrics are enabled by default when Trident is installed.

What you'll need

A Kubernetes cluster with Astra Trident installed.
A Prometheus instance. This can be a containerized Prometheus deployment or you can choose to run Prometheus as a native application.

Step 1: Define a Prometheus target

You should define a Prometheus target to gather the metrics and obtain information about the backends Astra Trident manages, the volumes it creates, and so on. This blog explains how you can use Prometheus and Grafana with Astra Trident to retrieve metrics. The blog explains how you can run Prometheus as an operator in your Kubernetes cluster and the creation of a ServiceMonitor to obtain Astra Trident’s metrics.

Step 2: Create a Prometheus ServiceMonitor

To consume the Trident metrics, you should create a Prometheus ServiceMonitor that watches the trident-csi service and listens on the metrics port. A sample ServiceMonitor looks like this:

apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
  name: trident-sm
  namespace: monitoring
  labels:
      release: prom-operator
  spec:
    jobLabel: trident
    selector:
      matchLabels:
        app: controller.csi.trident.netapp.io
    namespaceSelector:
      matchNames:
      - trident
    endpoints:
    - port: metrics
      interval: 15s

This ServiceMonitor definition retrieves metrics returned by the trident-csi service and specifically looks for the metrics endpoint of the service. As a result, Prometheus is now configured to understand Astra Trident's
metrics.

In addition to metrics available directly from Astra Trident, kubelet exposes many kubelet_volume_* metrics via it's own metrics endpoint. Kubelet can provide information about the volumes that are attached, and pods and other internal operations it handles. See here.

Step 3: Query Trident metrics with PromQL

PromQL is good for creating expressions that return time-series or tabular data.

Here are some PromQL queries that you can use:

Get Trident health information

Percentage of HTTP 2XX responses from Astra Trident

(sum (trident_rest_ops_seconds_total_count{status_code=~"2.."} OR on() vector(0)) / sum (trident_rest_ops_seconds_total_count)) * 100

Percentage of REST responses from Astra Trident via status code

(sum (trident_rest_ops_seconds_total_count) by (status_code)  / scalar (sum (trident_rest_ops_seconds_total_count))) * 100

Average duration in ms of operations performed by Astra Trident

sum by (operation) (trident_operation_duration_milliseconds_sum{success="true"}) / sum by (operation) (trident_operation_duration_milliseconds_count{success="true"})

Get Astra Trident usage information

Average volume size

trident_volume_allocated_bytes/trident_volume_count

Total volume space provisioned by each backend

sum (trident_volume_allocated_bytes) by (backend_uuid)

Get individual volume usage

This is enabled only if kubelet metrics are also gathered.

Percentage of used space for each volume

kubelet_volume_stats_used_bytes / kubelet_volume_stats_capacity_bytes * 100

Learn about Astra Trident AutoSupport telemetry

By default, Astra Trident sends Prometheus metrics and basic backend information to NetApp on a daily cadence.

To stop Astra Trident from sending Prometheus metrics and basic backend information to NetApp, pass the --silence-autosupport flag during Astra Trident installation.
Astra Trident can also send container logs to NetApp Support on-demand via tridentctl send autosupport. You will need to trigger Astra Trident to upload it's logs. Before you submit logs, you should accept NetApp's
privacy policy.
Unless specified, Astra Trident fetches the logs from the past 24 hours.
You can specify the log retention timeframe with the --since flag. For example: tridentctl send autosupport --since=1h. This information is collected and sent via a trident-autosupport container
that is installed alongside Astra Trident. You can obtain the container image at Trident AutoSupport.
Trident AutoSupport does not gather or transmit Personally Identifiable Information (PII) or Personal Information. It comes with a EULA that is not applicable to the Trident container image itself. You can learn more about NetApp's commitment to data security and trust here.

An example payload sent by Astra Trident looks like this:

{
  "items": [
    {
      "backendUUID": "ff3852e1-18a5-4df4-b2d3-f59f829627ed",
      "protocol": "file",
      "config": {
        "version": 1,
        "storageDriverName": "ontap-nas",
        "debug": false,
        "debugTraceFlags": null,
        "disableDelete": false,
        "serialNumbers": [
          "nwkvzfanek_SN"
        ],
        "limitVolumeSize": ""
      },
      "state": "online",
      "online": true
    }
  ]
}

The AutoSupport messages are sent to NetApp's AutoSupport endpoint. If you are using a private registry to store container images, you can use the --image-registry flag.
You can also configure proxy URLs by generating the installation YAML files. This can be done by using tridentctl install --generate-custom-yaml to create the YAML files and adding the --proxy-url argument for the trident-autosupport container in trident-deployment.yaml.

Disable Astra Trident metrics

To disable metrics from being reported, you should generate custom YAMLs (using the --generate-custom-yaml flag) and edit them to remove the --metrics flag from being invoked for the trident-main
container.