Create a Beyla DaemonSet manifest with the beyla container image, requesting the CAP_SYS_PTRACE and CAP_NET_ADMIN capabilities and a hostPID: true pod spec
Configure the BEYLA_OPEN_PORT or BEYLA_SERVICE_NAME environment variables to target specific processes or use BEYLA_DISCOVERY_SERVICES for process auto-discovery
Set the OTEL_EXPORTER_OTLP_ENDPOINT to point at the OTel Collector running as a DaemonSet or Deployment in the cluster
Create RBAC resources (ServiceAccount, ClusterRole, ClusterRoleBinding) granting Beyla access to node-level process information
Validate instrumentation by generating traffic and checking the Collector or backend for http.server.duration and rpc.server.duration metrics
Known gotchas
Beyla requires Linux kernel 5.8 or later for BTF-based CO-RE; older node kernel versions fall back to less stable non-BTF probes or fail entirely
hostPID: true gives the Beyla pod visibility into all host processes; this is a significant security boundary—restrict the DaemonSet to observability namespaces
eBPF programs are loaded per-process; services with many short-lived worker processes (forking servers) may cause excessive probe attachment overhead
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp