When we say auto-telemetry, we’re not talking about cars — we're talking about instant baseline visibility into your Kubernetes clusters. With the New Relic One integration with Pixie, you get similar data to traditional language agents, but without manually instrumenting your code or redeploying your application.
Pixie auto-telemetry is powered by eBPF, a virtual machine-like construct that enables Pixie to seamlessly collect fine-grained telemetry data — service-level metrics, unsampled requests, and more. With one install command, you get deeper insight into your Kubernetes clusters and workloads. No language agents required.
Simply put, Auto-telemetry with Pixie offers the quickest option for getting observability into your Kubernetes services.
Our Pixie integration gives you the best of both worlds: Pixie’s fast and simple Kubernetes observability coupled with New Relic One’s incident correlation, intelligent alerting, and long-term retention.
You’ll get visibility into HTTP services using golden signals, HTTP transactions, database transactions, distributed tracing, and JVM metrics. You can operate, debug, and scale your Kubernetes clusters based on the information you learn about how your clusters and services are running. Using the New Relic Explorer, you can see key metrics and events at every level, starting with the cluster, and diving down into namespaces, deployments, and pods. You can quickly spot anomalous behavior, and where it’s happening.
And then dive deeper using embedded visualizations of your Pixie data. Quickly identify hot spots with Flamegraph. On the Live debugging with Pixie tab, answer questions like what SQL requests your app is making or which services are talking to each other.
Use our guided installation process to install Auto-telemetry with Pixie. This deploys Pixie with New Relic's Kubernetes integration on your cluster. You don't need to do any further configuration or installation to start using Pixie.
If you want to install Auto-telemetry with Pixie on multiple clusters, re-run the guided install for each additional cluster.
- Review this Pixie data security overview for actions to take to secure your data.
- Make sure you have sufficient memory: Pixie requires 2Gb of memory per node in your cluster.
If you are already a Pixie user, you must still install using the guided installation steps described below. This will provide the API keys that you need.
- Open our New Relic One guided install.
- Select the account you want to use for the guided install, and click Continue. Note: if you have a single account, you won't see this option.
- Select Kubernetes and then continue with step one in the next section.
If you arrived in the guided installation process by following a link from Pixie or from within New Relic, your steps begin here.
Select the account and cluster for the install. If needed, select a namespace.
Currently, Pixie performs best on clusters with up to 100 nodes (exceeding 100 nodes can lead to excessive memory usage and scripts failing to run). Friendly reminder: autoscaling can quickly drive up your node numbers.
Select the data you want to gather, observe, and debug, and click Continue.
On the Choose install method page, copy the Helm command that's provided, and then run it on your command line. See this page about installing the Kubernetes integration using Helm to learn more about the process.
Helm installs a bundle containing the New Relic infrastructure agent, an integration to gather Prometheus metrics and Kubernetes events, and the Pixie integration. The deployment takes a few minutes to complete.
To see the status of the install to the cluster, run
kubectl get pods -n newrelic.
Click Continue to open the Listening for data page.
When you get the message, See your data, click Kubernetes Cluster Explorer to see your cluster.
Auto-telemetry with Pixie might restart after installation. This is caused by the auto update feature.
In the cluster explorer, you can get a quick overview of the nodes in your cluster, including CPU, memory, and storage, as well as the status of each pod (healthy, warning, or critical). You can also find out what services are running in each container, their latency, throughput, and error rate.
For more information about using the cluster explorer, see Navigate the Kubernetes cluster explorer.
Containers might be listed for up to four hours after they get decommissioned.
You can query the Pixie data in New Relic One and create dashboards for at-a-glance monitoring. Find the data model and sample queries here.
Debugging is orders of magnitude easier when you can quickly see what your application is doing. Flamegraph, a Pixie always-on visualization, requires no instrumentation, redeploying, or recompiling. It works for compiled languages like Go, C+, Rust, to name a few. And at a glance, Flamegraph tells you what functions your application is spending time on and where you have hot spots. Flamegraph is especially useful for hierarchical resource use, like disk usage and CPU utilization. For more information on how to read Flamegraph, see the Pixie documentation.
On the Live debugging with Pixie tab, run PxL scripts — scripts written in Pixie's PxL language — to view live data captured through eBPF. Select the script drop-down and then select a script to run in the tab. (For best results, select a time range that is recent in the time picker.)
Scripts enable you to debug:
- Traffic in multiple formats: HTTP and HTTPs (including encrypted), DNS, Postgres, MySQL, Cassandra, Redis (currently supporting SQL and HTTP in beta)
- Kubernetes services and their throughput, error rate, and latency statistics
- Service maps to learn which services are talking to each other
- Network traffic maps to learn which nodes are talking to each other
- JVM data