Install the Chronosphere Collector

The Chronosphere Collector can either collect metrics data or receive trace data. When collecting metrics data, the Collector pulls data it scrapes from Prometheus endpoints. When receiving trace data, an app pushes data to a service, with the Collector behind that service receiving the data.

Regardless of the installation method you use, Chronosphere recommends deploying separate, independent Collectors for metrics and for traces. This distinction means that a Collector either collects metrics data or receives trace data.

Collect metrics data

You can install the Collector to retrieve metrics data, either with Kubernetes or as a standalone binary. When deployed within Kubernetes, the Collector scrapes Prometheus endpoints exposed on pods, based on pod-level annotations.

With Kubernetes

You can install the Collector with Kubernetes to collect metrics as:

Chronosphere recommends installing the Collector as a Kubernetes DaemonSet on each node in a cluster to:

  • Minimize cross-node network traffic within a cluster.
  • Eliminate a single point of failure in case one node is offline.
  • Provide the ability to switch to a standalone deployment if the Collector is using too many resources.

Without Kubernetes

If you don't use Kubernetes, or want to gather metrics from services not managed by Kubernetes, you can install the Collector as a standalone binary.

Docker images for versions 0.93.0 and later are multiplatform builds for both linux/amd64 and linux/arm64. Chronosphere also offers Linux binaries for both of these platforms.

Receive trace data

To receive trace data, Chronosphere recommends installing the Collector as a Kubernetes Deployment, typically with a service in front to receive trace data. Your tracing app pushes trace data to the service, which the Collector receives the data from.