Pumba
github.com
What it is
Pumba is a free, open-source chaos testing and network emulation tool for Docker, containerd, and Podman containers. It can kill, stop, pause, remove, and restart containers, while network experiments can add delay, packet loss, duplication, corruption, or rate limits through netem or iptables. Its stress command exercises CPU, memory, and I/O with stress-ng. Targets can be chosen by container name, regular expression, label, or random selection, and experiments can recur at fixed intervals. Pumba documents deployment with Docker, Kubernetes DaemonSets, and OpenShift, and supports Kubernetes workflows. Released binaries and a GitHub Container Registry image are available; the Docker Hub image is marked deprecated. Linux is the primary platform. macOS binaries drive a remote Linux container VM, while Windows is not supported or planned. Pumba runs self-hosted under the Apache License 2.0 and requires privileged access to container runtimes, so its guidance calls for controlled environments and restricted socket access. Security fixes apply only to the latest released minor version. Community support is directed to GitHub Issues and the project blog.
Spec sheet
- Purpose
- Pumba is a chaos testing and network emulation tool for Docker, containerd, and Podman containers.Read 8 Oct 2026 · github.com
- Deployment
- The project documents Docker, Kubernetes DaemonSets, and OpenShift deployment options.Read 8 Oct 2026 · github.com
- Support
- The README points users to GitHub Issues and a project blog for community support.Read 8 Oct 2026 · github.com
- Container disruptions
- It can kill, stop, pause, remove, and restart containers.Read 8 Oct 2026 · github.com
- Distribution
- Pumba is available as released binaries and a GitHub Container Registry image; its README marks the Docker Hub image as deprecated.Read 8 Oct 2026 · github.com
- Network testing
- It can inject network delay, packet loss, duplication, corruption, and rate limits using netem or iptables.Read 8 Oct 2026 · github.com
- Runtimes
- Pumba supports Docker, containerd, and Podman, with containerd and Podman-specific socket and privilege requirements.Read 8 Oct 2026 · github.com
- Security requirements
- Pumba interacts directly with container runtimes and requires privileged access; its security guidance recommends controlled environments and restricting socket access.Read 8 Oct 2026 · github.com
- Security support
- Security fixes apply only to the latest released minor version, and vulnerabilities should be reported privately through GitHub or to the maintainer.Read 8 Oct 2026 · github.com
- Security tooling
- The repository says it uses CodeQL, golangci-lint with gosec, and GitHub Actions with least-privilege permissions.Read 8 Oct 2026 · github.com
- Stress testing
- Its stress command tests CPU, memory, and I/O using stress-ng.Read 8 Oct 2026 · github.com
- Supported platforms
- Linux is the primary target, while macOS binaries are for driving a remote Linux container VM; Windows is not supported or planned.Read 8 Oct 2026 · github.com
- Targeting and scheduling
- Targets can be selected by names, regular expressions, labels, or randomly, and chaos can recur at fixed intervals.Read 8 Oct 2026 · github.com
Plans and prices
Strong points and limits
Strong points
- Free open-source Apache 2.0 license
- Supports Docker, containerd, and Podman
- Injects several network fault types
- Schedules recurring chaos experiments
- Supports Kubernetes and OpenShift deployment
Limits
- Windows is not supported or planned
- Requires privileged runtime access
- Cloud fault injection is unavailable
- Docker Hub image is deprecated
Who it suits
It suits teams running containerized systems that need self-hosted failure injection, network emulation, stress tests, and scheduled experiments. Linux is the main target; macOS use depends on a remote Linux VM.
Verdict
Pumba provides broad container disruption, network fault, and stress controls in a self-hosted package. Its Linux focus and privileged runtime requirements make deployment security and platform choice central considerations.
Where Specifiction places it
- Chaos Engineering Platforms#18 of 257.3