To set up eBPF profiling with OpsRamp on Kubernetes, you need to:
- Verify that your cluster meets the Prerequisites.
- Add the OpsRamp Agent helm repository.
There are two ways to enable eBPF Network Visibility:
- Method 1: Enable During Kubernetes 2.0 Agent Installation
- Method 2: Enable via Helm Upgrade
Method 1: Enable During Kubernetes 2.0 Agent Installation
- Follow the standard Kubernetes 2.0 Agent Installation Instructions.
- Use the Helm install command with eBPF enabled:
helm install opsramp oci://us-docker.pkg.dev/opsramp-registry/agent-helm-charts/agent -n opsramp-agent --create-namespace --values <download-file-path-or-copied-file-path>.yaml --set ebpf.enabled=true --set ebpf.logEndpoint="int.opsramp.net:443" --set agent.config.containerEngine="<cri-o|containerd>" - Optional: Specify custom container runtime socket paths:
--set containerSockFile.containerd=/path/custom/containerd.sock --set containerSockFile.cri-o=/path/custom/crio.sock
Important values to set:
| Parameter | Description | Required/Optional | Default Value |
|---|---|---|---|
| ebpf.enabled | Enable eBPF feature | Optional | false |
| ebpf.logEndpoint | OpsRamp eBPF log endpoint | Mandatory if enabled | e.g., int.opsramp.net:443 |
| agent.config.containerEngine | Container runtime: "cri-o" or "containerd" | Mandatory if enabled | - |
| containerSockFile.containerd | Socket path for containerd runtime | Optional | /var/run/containerd/containerd.sock |
| containerSockFile.cri-o | Socket path for cri-o runtime | Optional | /var/run/crio/crio.sock |
| ebpf.podCIDRs | List of pod CIDRs treated as internal traffic | Optional | - |
Method 2: Enable via Helm Upgrade
For clusters with an existing OpsRamp Kubernetes 2.0 Agent:
helm upgrade oci://us-docker.pkg.dev/opsramp-registry/agent-helm-charts/agent -n opsramp-agent --create-namespace --values <download-file-path-or-copied-file-path>.yaml --set ebpf.enabled=true --set ebpf.logEndpoint="int.opsramp.net:443" Troubleshooting
If you encounter an issue, Validate data collection and export by checking the logs of the eworker pod.