- Quick Start
- Custom Filters
- Deploying to Kubernetes from GitHub
- Knative Serverless Framework
- Prometheus monitoring
- The Mapping Resource
- Automatic Retries
- Canary Releases
- Circuit Breakers
- Cross-Origin Resource Sharing
- Method-Based Routing
- Prefix Regex
- Query Parameter-Based Routing
- Traffic Shadowing
- The Ambassador Module
- Custom Error Responses
- Gzip Compression
- Host CRD, ACME Support, and External Load Balancer Configuration
- Ingress Controller
- Scaling Ambassador
- Developer Portal
Traffic shadowing is a deployment pattern where production traffic is asynchronously copied to a non-production service for testing. Shadowing is a close cousin to two other commonly known deployment patterns, canary releases and blue/green deployments. Shadowing traffic has several important benefits over blue/green and canary testing:
Zero production impact. Since traffic is duplicated, any bugs in services that are processing shadow data have no impact on production.
Test persistent services. Since there is no production impact, shadowing provides a powerful technique to test persistent services. You can configure your test service to store data in a test database, and shadow traffic to your test service for testing. Both blue/green deployments and canary deployments require more machinery for testing.
Test the actual behavior of a service. When used in conjunction with tools such as Twitter's Diffy, shadowing lets you measure the behavior of your service and compare it with an expected output. A typical canary rollout catches exceptions (e.g., HTTP 500 errors), but what happens when your service has a logic error and is not returning an exception?
Ambassador Edge Stack lets you easily shadow traffic to a given endpoint. In Ambassador Edge Stack, only requests are shadowed; responses from a service are dropped. All normal metrics are collected for the shadow services. This makes it easy to compare the performance of the shadow service versus the production service on the same data set. Ambassador Edge Stack also prioritizes the production path, i.e., it will return responses from the production service without waiting for any responses from the shadow service.
In Ambassador Edge Stack, you can enable shadowing for a given mapping by setting
shadow: true in your
Mapping. One copy proceeds as if the shadowing
Mapping was not present: the request is handed onward per the
service(s) defined by the non-shadow
Mappings, and the reply from whichever
service is picked is handed back to the client.
The second copy is handed to the
service defined by the
shadow set. Any reply from this
service is ignored, rather than being handed back to the client. Only a single
shadow per resource can be specified (i.e., you can't shadow the same resource to more than 1 additional destination). In this situation, Ambassador Edge Stack will indicate an error in the diagnostic service, and only one
shadow will be used. If you need to implement this type of use case, you should shadow traffic to a multicast proxy (or equivalent).
You can shadow multiple different services.
During shadowing, the host header is modified such that
-shadow is appended.
The following example may help illustrate how shadowing can be used. This first attribute sets up a basic mapping between the
myservice Kubernetes service and the
/myservice/ prefix, as expected.
---apiVersion: getambassador.io/v2kind: Mappingmetadata:name: myservicespec:prefix: /myservice/service: myservice.default
What if we want to shadow the traffic to
myservice, and send that exact same traffic to
myservice-shadow? We can create a new mapping that does this:
---apiVersion: getambassador.io/v2kind: Mappingmetadata:name: myservice-shadowspec:prefix: /myservice/service: myservice-shadow.defaultshadow: true
prefix is set to be the same as the first mapping, which tells Ambassador Edge Stack which production traffic to shadow. The destination service, where the shadow traffic is routed, is a different Kubernetes service,
myservice-shadow. Finally, the
shadow: true attribute actually enables shadowing.
It is possible to shadow a portion of the traffic by specifying the
weight in the mapping. For example:
---apiVersion: getambassador.io/v2kind: Mappingmetadata:name: myservice-shaddowspec:prefix: /myservice/service: myservice-shadow.defaultshadow: trueweight: 10
In the example above, only 10% of the traffic will be forwarded to the shadowing service.