AWS X-Ray
Trace requests through distributed and serverless apps on AWS.
Zipkin
Trace, visualize, and troubleshoot latency across microservices
Zipkin collects distributed tracing data, lets you query by service, operation, tags, or duration, and visualizes request flows with a UI and dependency diagram.
Overview
Overview
Zipkin is a distributed tracing system that gathers timing data from instrumented services, enabling developers to query traces by service name, operation, tags, or duration. The built‑in UI presents trace details and a dependency graph that highlights request paths and error flows, helping teams pinpoint latency bottlenecks.
Deployment & Storage
Zipkin can be started instantly via a self‑contained JAR or the official Docker image. It supports multiple storage backends—including in‑memory for testing, Cassandra, and Elasticsearch—for persistent span storage. Reporting transports such as HTTP, Kafka, gRPC, RabbitMQ, and others allow flexible integration with existing pipelines. A slimmer distribution is available for faster startup when only in‑memory or Elasticsearch storage is needed.
Instrumentation
Applications must be instrumented with a Zipkin tracer or a compatible library (e.g., Brave). Once spans are emitted, Zipkin aggregates, indexes, and serves them for interactive exploration or programmatic retrieval.
Highlights
Searchable trace data by service, operation, tags, and duration
Dependency graph visualizing inter‑service calls
Multiple storage backends (in‑memory, Cassandra, Elasticsearch) and transport options (HTTP, Kafka, gRPC, etc.)
Slim distribution for fast startup and minimal footprint
Pros
- Rich query capabilities across trace attributes
- UI with dependency diagram aids root‑cause analysis
- Supports many storage and transport integrations
- Easy quick‑start via Docker or self‑contained JAR
Considerations
- Requires instrumentation of each service
- Full server needs Java 17+, increasing runtime overhead
- In‑memory storage is not suitable for production workloads
- Disabling search reduces discoverability of traces
Managed products teams compare with
When teams consider Zipkin, these hosted platforms usually appear on the same shortlist.
Looking for a hosted option? These are the services engineering teams benchmark against before choosing open source.
Fit guide
Great for
- Teams needing end‑to‑end latency visibility in microservice architectures
- Organizations that already use Java‑based services and want a lightweight tracing solution
- Environments where a dependency graph helps identify error paths
- Projects that can provision Cassandra or Elasticsearch for persistent storage
Not ideal when
- Applications without tracing instrumentation or unwilling to modify code
- Deployments limited to low‑memory environments where even the slim build is too large
- Use cases requiring out‑of‑the‑box support for proprietary messaging systems not listed
- Teams that need a fully managed SaaS tracing service
How teams use it
Debugging high latency in a payment service
Identify slow downstream calls and visualize the call chain to pinpoint bottlenecks.
Monitoring service degradation after a deployment
Query recent traces by tag to see error rates and view affected services in the dependency diagram.
Testing new microservice locally
Run Zipkin with in‑memory storage to capture traces without external databases.
Integrating with existing Kafka pipeline
Configure services to send spans via Kafka and store them in Elasticsearch for long‑term analysis.
Tech snapshot
Java78%
TypeScript9%
JavaScript9%
Shell2%
Dockerfile1%
HTML1%
Frequently asked questions
What Java version is required to run the Zipkin server?
The server requires at least JRE 17; the core library works with Java 8+.
Can I use Zipkin without a persistent storage backend?
Yes, the in‑memory storage works for testing, but it is not suitable for production workloads.
How do I start Zipkin quickly?
Download the self‑contained JAR via the quick‑start script or run the official Docker image on port 9411.
What transport protocols are supported for sending spans?
Spans can be reported over HTTP, Kafka, gRPC, RabbitMQ, Apache Pulsar, ActiveMQ, among others.
How can I disable the search API?
Set the environment variable SEARCH_ENABLED=false or configure StorageComponent.Builder.searchEnabled(false).
Project at a glance
Stable- Stars
- 17,396
- Watchers
- 17,396
- Forks
- 3,105
LicenseApache-2.0
Repo age13 years old
Last commit5 months ago
Primary languageJava
Last synced yesterday