zipkin-gcp

Shared libraries that provide Zipkin integration with the Google Cloud Platform. Requires JRE 17 or later.

Usage

These components integrate traced applications and servers through Google Cloud services via interfaces defined in Zipkin and zipkin-reporter-java.

Senders

The component in a traced application that sends timing data (spans) out of process is called a Sender. Senders are called on interval by an async reporter.

NOTE: Applications can be written in any language, while we currently only have senders in Java, senders in other languages are welcome.

Encoders

Encoding is library-specific, as some libraries use zipkin2.Span and others brave.handler.MutableSpan. Both options are available to encode to the StackDriver Trave V2 format.

Collectors

The component in a zipkin server that receives trace data is called a collector. This decodes spans reported by applications and persists them to a configured storage component.

Storage

The component in a zipkin server that persists and queries collected data is called StorageComponent. This primarily supports the Zipkin Api and all collector components.

Server integration

In order to integrate with zipkin-server, you need to use properties launcher to load your collector (or sender) alongside the zipkin-server process.

To integrate a module with a Zipkin server, you need to:

• add a module jar to the loader.path
• enable the profile associated with that module
• launch Zipkin with PropertiesLauncher

Each module will also have different minimum variables that need to be set.

Ex.

$ curl -sSL https://zipkin.io/quickstart.sh | bash -s
$ curl -sSL https://zipkin.io/quickstart.sh | bash -s io.zipkin.gcp:zipkin-module-gcp:LATEST:module gcp.jar
$ STORAGE_TYPE=stackdriver STACKDRIVER_PROJECT_ID=zipkin-demo \
    java \
    -Dloader.path='gcp.jar,gcp.jar!/lib' \
    -Dspring.profiles.active=gcp \
    -cp zipkin.jar \
    org.springframework.boot.loader.launch.PropertiesLauncher

Example integrating Stackdriver Storage

If you cannot use our Docker image, you can still integrate yourself by downloading a couple jars.

Here's an example of integrating Stackdriver storage.

Troubleshooting

Translation issues

When using a component that sends data to Stackdriver, if you see nothing in the console, try enabling DEBUG logging on the translation component. If using Spring Boot (ex normal app or zipkin server integration), add the following system property:

-Dlogging.level.zipkin2=DEBUG

Note: If using our docker image or anything that uses JAVA_OPTS, you can add this there.

With this in place, you'll see the input and output of translation like below. Keep a copy of this when contacting us on gitter for support.

2018-04-09 13:58:44.112 DEBUG [/] 11325 --- [   XNIO-2 I/O-3] z.t.stackdriver.SpanTranslator           : >> translating zipkin span: {"traceId":"d42316227862f939","parentId":"d42316227862f939","id":"dfbb21f9cf4c52b3","kind":"CLIENT","name":"get","timestamp":1523253523054380,"duration":4536,"localEndpoint":{"serviceName":"frontend","ipv4":"192.168.1.113"},"tags":{"http.method":"GET","http.path":"/api"}}
2018-04-09 13:58:44.113 DEBUG [/] 11325 --- [   XNIO-2 I/O-3] z.t.stackdriver.SpanTranslator           : << translated to stackdriver span: span_id: 16199746076534411288
kind: RPC_CLIENT
name: "get"
start_time {
  seconds: 1523253523
  nanos: 54380000
}
end_time {
  seconds: 1523253523
  nanos: 58916000
}
parent_span_id: 15286085897530046777
labels {
  key: "/http/method"
  value: "GET"
}
labels {
  key: "zipkin.io/http.path"
  value: "/api"
}
labels {
  key: "/component"
  value: "frontend"
}

Healthcheck API

If you are running Zipkin server, /health HTTP endpoint can be used to check service health.

gRPC Headers

If you suspect an issue between Zipkin server and Stackdriver, inspecting gRPC headers may be useful. Set STACKDRIVER_HTTP_LOGGING environment variable to HEADERS to log gRPC status information.

GCP console

If you believe that spans are reaching Stackdriver, verify what happens to them from the GCP side by visiting APIs & Services > Dashboard > Stackdriver Trace API > Metrics section.

• Is there any traffic at all? If not, then the requests are likely not reaching Stackdriver at all. Check your applications/proxies to make sure requests go to the right place.
• Are there errors? If so, narrow down the source of errors by selecting specific Credentials and Methods values from the filter drop-downs on top and seeing the effect in "Errors by API method" chart or "Methods" table.
  • If all writes are failing, check that your service account has access to cloudtrace.agent role or cloudtrace.traces.patch permission.
  • If reads are failing, check for cloudtrace.user role or the specific permission (full permissions list). NOTE: read and write permissions are separate; only the admin role has both.

Artifacts

All artifacts publish to the group ID "io.zipkin.gcp". We use a common release version for all components.

Library Releases

Releases are at Sonatype and Maven Central

Library Snapshots

Snapshots are uploaded to Sonatype after commits to master.

Docker Images

Released versions of zipkin-gcp are published to Docker Hub as openzipkin/zipkin-gcp and GitHub Container Registry as ghcr.io/openzipkin/zipkin-gcp.

See docker for details.