A lightweight eBPF-based tool that instantly visualizes Kubernetes service dependencies and network traffic in Grafana.
Caretta is an open-source tool that automatically generates a real-time network dependency map for Kubernetes services. It uses eBPF to trace all inter-service communications without code instrumentation, then visualizes the data in Grafana to help teams understand traffic flows and identify bottlenecks. It solves the problem of opaque service dependencies in complex microservices architectures.
Platform engineers, SREs, and DevOps practitioners managing Kubernetes clusters who need visibility into service dependencies and network traffic patterns.
Developers choose Caretta because it provides instant, zero-configuration dependency mapping using lightweight eBPF technology, integrates seamlessly with existing Grafana setups, and requires no application changes—making it ideal for production environments.
Instant K8s service dependency map, right to your Grafana.
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Leverages extended Berkeley Packet Filter for tracing network sockets with minimal performance overhead, as highlighted in the Key Features, ensuring low resource usage without cluster modifications.
Installs via a simple Helm chart with a small footprint and no need to modify existing workloads, making it quick to deploy in Kubernetes clusters.
Automatically maps pods, services, and deployments to their owners for meaningful entity names, enhancing dependency visualization without manual configuration.
Ships with a preconfigured Grafana dashboard and exports Prometheus-compatible metrics, allowing seamless integration with existing observability stacks for custom queries.
Requires Linux kernel >= 4.16 and CO-RE support, excluding popular development environments like Docker for Mac and older distributions, as admitted in the Requirements section.
Focuses on network link observations (e.g., bytes transferred via caretta_links_observed) without deeper insights such as latency percentiles or application-layer errors, limiting troubleshooting depth.
If persistent volume is enabled, uninstallation may not automatically delete it, potentially leaving orphaned resources and requiring manual cleanup, as noted in the Uninstallation instructions.