We can access Grafana directly, or we can jump to context-based access from the Linkerd dashboard by clicking the Grafana icon. Switch back to the Linkerd dashboard and follow these steps:
- Go to Resources | Namespaces | Linkerd-Lab:
The Linkerd dashboard for the namespace shows the service dependency diagram and shows metrics based on deployment, pods, authorities, and TCP.
- Click the Grafana icon under Deployment/Authors:
Notice the graphical view of the success rate, request rate per second (RPS), inbound and outbound deployments, latency, and so on.
- Scroll down to view the deploy/books in the Grafana dashboard. Alternatively, you can select it from Linkerd's dashboard dropdown:
- From the left sidebar, click on the dashboard (the four-square icon) and click Home.
This panel shows the metrics for several namespaces that Linkerd is managing, along with the total number of deployments and the overall success rate. The global latency metrics are useful for drilling down into latency issues downstream:
- Scroll down and check for emojivoto and the Linkerd control plane at the namespace level:
- Scroll down to the linkerd-lab booksapp namespace. This shows you the overall metrics at the namespace level:
If you click ns/linkerd-lab, then you can drill down to the deployment and pod level to look at the metrics. The visualization of metrics in Grafana is collected through Prometheus, which puts observability into the service mesh and makes things easier in terms of operational purposes. The visual representation of the success rate helps us close the feedback loop with the development team so that they can improve upon the issues that were detected through observability.
Linkerd only provides 6 hours of data collection in order to keep the footprint small. To deal with this and maintain the long-term storage of data, we need to look at external Prometheus integration. We'll do this in the next section.