// the find
apache/skywalking
APM, Application Performance Monitoring System
SkyWalking is a full-stack APM platform built for microservices and Kubernetes — distributed tracing, metrics, logs, and continuous profiling in one system. It ships its own database (BanyanDB), supports a wide range of ingestion formats (OpenTelemetry, Zipkin, Prometheus, Zabbix), and has agents for most major languages. The target audience is platform/SRE teams running polyglot microservices at scale who want a self-hosted alternative to Datadog or New Relic.
Active development with a v10.x release line and daily commits; the eBPF Rover agent gives you CPU and network profiling on Kubernetes without code changes. BanyanDB as a dedicated observability storage backend avoids the Elasticsearch dependency that made older versions painful to operate. The OAL/MAL/LAL scripting pipeline lets you define custom aggregation, log parsing, and alerting rules without forking the server — a meaningful operational advantage over most open-source APMs. Multi-signal correlation (trace → metric → log → profiling) is built into the data model, not bolted on afterward.
The configuration surface is enormous — application.yml has hundreds of keys, and wrong combinations fail silently or with cryptic startup errors. BanyanDB is immature; the docs warn you to treat it as 'production-ready with caveats', and most real deployments still default to Elasticsearch or Banyan with extra caution. The Java agent is the first-class citizen; the .NET, Go, and Python agents lag significantly in feature coverage and are community-maintained with variable quality. Setup overhead is high — a minimal useful deployment (OAP server + UI + storage + at least one agent) takes meaningful infrastructure effort before you see a single trace.