3. Deployments

This section describes Observability Framework deployment options.

3.1. Collector Service Deployments

3.1.1. Current Release

This release supports running Collector Service in storage mode only. Data can be stored into Prometheus and Grafana Loki natively or into other third-party observability backends using OpenTelemetry and the OpenTelemetry Collector. Because forwarding mode is not supported, you can only use a single layer of Collector Services per Connext system. This configuration is illustrated in Figure 3.1 and Figure 3.2

Figure 3.1 Single Layer Collector Deployment

Figure 3.2 Single Layer Collector Deployment using OpenTelemetry Collector

The deployments represented in Figure 3.1 and Figure 3.2 require running multiple instances of Collector Service where each Connext application configures Monitoring Library 2.0 to connect to one of the Collector Service instances. This release distributes Collector Service as a Docker image; for details, refer to the Docker (Separate Deployment) section.

You are responsible for running the Collector Service instances and the third-party components for storage. For example, if you want to store telemetry data into Prometheus and Grafana Loki, you must run Prometheus and Grafana Loki instances and configure the Docker container for Collector Service to connect to these storage backends.

Alternately, you can use a single layer deployment to run only one Collector Service instance for the Connext system, as illustrated in Figure 3.3. To deploy Observability Framework using a single collector (including the third-party components Prometheus and Grafana), use Docker Compose as described in the Docker Compose (Prepackaged) section.

Figure 3.3 Single Collector Deployment

3.1.2. Future Releases

As you roll out telemetry data collection and distribution across all your Connext applications, Observability Framework must be deployed in a way that supports the additional load. A single layer Collector Service deployment, as shown in Current Release, may not scale sufficiently.

A better deployment option would be the layered deployment depicted in Figure 3.4 and Figure 3.5. In this option, you have multiple layers of Collector Service gathering, filtering, and forwarding the telemetry data produced by the Connext applications. Each intermediate layer reduces the number of egress points required to send data and provides an opportunity to filter telemetry data. The last layer works as a storage layer and is responsible for storing the telemetry data into a third-party observability backend.

Figure 3.4 Layered Collector Deployment

Figure 3.5 Layered Collector Deployment Using OpenTelemetry Collector