WatchLog
Open-source monitoring for Kubernetes and Docker. Install one agent and see every node, pod and container, and the network between your services, in about five minutes. No PromQL, no dashboards to build.
What you get
- One install. A Helm chart for Kubernetes, or one
docker run per Docker host.
- Ready-made views. Cluster overview, nodes, workloads, pods with restart and OOM reasons, events, and Docker hosts and containers. Press ⌘K to jump to any of them by name.
- Why it broke. Every alert and problem gets its probable cause in one sentence, with the evidence behind it: "web fails because api was scaled to 0: nothing answers its calls." Twelve fixed rules over what WatchLog already collects, across the service map, on Kubernetes and Docker; the same line in Slack, Telegram and email. No AI in the product, no scores; see probable cause.
- A service map without code changes. eBPF shows who connects to whom, failed connections, connect latency, bytes and DNS errors. External services are named by DNS.
- Your databases, caches, queues and web servers. Redis, PostgreSQL, MySQL and MariaDB, MongoDB, Nginx, Traefik, RabbitMQ and Kafka are found by themselves; press Monitor for their own page, the reasons behind a problem (the session holding a row lock, the consumer group that lost its members, the upstream sending 5xx) and default alerts. Read-only access, and HTTP stats from eBPF with nothing to set up.
- Logs when you need them, and sizes that fit. A container's last lines on its page, streamed and never stored (logs on demand). Requests and limits compared with what each container really uses, in plain sentences (right-sizing).
- Alerts that work on day one. Crash loops, OOM kills, node pressure, failing connections and more, sent to Slack, Telegram, email or webhooks.
- A small, read-only agent. Under 64 MB of memory and 1 % of a core. It never reads Secrets, ConfigMaps, environment values or payloads; see what it can access.
- For your team. Owners, Admins and Viewers, invited by email; an MCP server and a REST API for your AI assistant and scripts, read-only.
- Self-hosted and free, for as many Kubernetes clusters, Docker fleets and nodes as you run, under AGPL-3.0.
Install in five minutes
Kubernetes
helm install watchlog oci://ghcr.io/watchlog-light/charts/watchlog --version 0.5.0 \
--namespace watchlog --create-namespace
kubectl -n watchlog port-forward svc/watchlog-web 8080
Open http://localhost:8080 and enter the setup code from kubectl -n watchlog logs deploy/watchlog-backend -c backend | grep "setup code". The cluster appears within a minute.
Docker hosts
mkdir watchlog && cd watchlog
curl -fsSLO https://raw.githubusercontent.com/watchlog-light/watchlog/v0.5.0/deploy/docker/compose.yaml
curl -fsSL -o .env https://raw.githubusercontent.com/watchlog-light/watchlog/v0.5.0/deploy/docker/.env.example
# set WATCHLOG_DB_PASSWORD, WATCHLOG_SECRET_KEY (openssl rand -base64 32), WATCHLOG_PUBLIC_URL in .env
docker compose up -d && docker compose logs backend | grep "setup code"
Then add your hosts from Settings → Clusters: it shows the docker run command for each host.
The install guide covers requirements, PodSecurity, more clusters, upgrades (never with manual steps) and uninstalling.
Documentation
Repository layout
| Path | What |
|---|
agent/ | Go agent (node, cluster and Docker modes) and eBPF code |
backend/ | Node.js 24 + TypeScript + Fastify API and alert worker |
web/ | Vue 3 + Vite dashboard |
packages/design-tokens/ | Colour tokens, Tailwind theme, logo |
packages/api-types/ | API types shared with the dashboard |
ee/ | Enterprise modules (commercial licence) |
deploy/ | Docker Compose files, Helm charts and end-to-end tests |
docs/ | Spec, decisions, security, user docs and release notes |
Development
Prerequisites: Node.js 24 (.nvmrc), corepack (corepack enable provides the pinned pnpm), Go 1.27, Docker, GNU Make.
This is the beginning of the README. Read the rest at the original. Read the full README ↗
