Cache filter

In this example, we demonstrate how HTTP caching can be utilized in Envoy by using the Cache Filter. The setup of this sandbox is based on the setup of the Front Proxy sandbox.

All incoming requests are routed via the front Envoy, which acts as a reverse proxy sitting on the edge of the envoymesh network.

Port 8000 is exposed by docker-compose.yaml to handle HTTP calls to the services. Two backend services are deployed behind the front Envoy, each with a sidecar Envoy.

The front Envoy is configured to run the Cache Filter, which stores cacheable responses in an in-memory cache, and serves it to subsequent requests.

In this demo, the responses that are served by the deployed services are stored in responses.yaml.

This file is mounted to both services’ containers, so any changes made to the stored responses while the services are running should be instantly effective (no need to rebuild or rerun).

For the purposes of the demo, a response’s date of creation is appended to its body before being served. An Etag is computed for every response for validation purposes, which only depends on the response body in the yaml file (i.e. the appended date is not taken into account). Cached responses can be identified by having an age header. Validated responses can be identified by having a generation date older than the date header; as when a response is validated the date header is updated, while the body stays the same. Validated responses do not have an age header. Responses served from the backend service have no age header, and their date header is the same as their generation date.

Step 1: Start all of our containers

Change to the examples/cache directory.

$ pwd
envoy/examples/cache
$ docker-compose build --pull
$ docker-compose up -d
$ docker-compose ps

       Name                      Command            State           Ports
----------------------------------------------------------------------------------------------
cache_front-envoy_1   /docker-entrypoint.sh /bin ... Up      10000/tcp, 0.0.0.0:8000->8000/tcp
cache_service1_1      /bin/sh -c /usr/local/bin/ ... Up      10000/tcp, 8000/tcp
cache_service2_1      /bin/sh -c /usr/local/bin/ ... Up      10000/tcp, 8000/tcp

Step 2: Test Envoy’s HTTP caching capabilities

You can now send a request to both services via the front-envoy. Note that since the two services have different routes, identical requests to different services have different cache entries (i.e. a request sent to service 2 will not be served by a cached response produced by service 1).

To send a request:

curl -i localhost:8000/service/<service_no>/<response>

service_no: The service to send the request to, 1 or 2.

response: The response that is being requested. The responses are found in responses.yaml.

The provided example responses are:

  • valid-for-minute

    This response remains fresh in the cache for a minute. After which, the response gets validated by the backend service before being served from the cache. If found to be updated, the new response is served (and cached). Otherwise, the cached response is served and refreshed.

  • private

    This response is private; it cannot be stored by shared caches (such as proxies). It will always be served from the backend service.

  • no-cache

    This response has to be validated every time before being served.

You can change the responses’ headers and bodies (or add new ones) while the sandbox is running to experiment.

Example responses

1. valid-for-minute

$ curl -i localhost:8000/service/1/valid-for-minute
HTTP/1.1 200 OK
content-type: text/html; charset=utf-8
content-length: 103
cache-control: max-age=60
custom-header: any value
etag: "172ae25df822c3299cf2248694b4ce23"
date: Fri, 11 Sep 2020 03:20:40 GMT
server: envoy
x-envoy-upstream-service-time: 11

This response will stay fresh for one minute
Response body generated at: Fri, 11 Sep 2020 03:20:40 GMT

Naturally, response date header is the same time as the generated time. Sending the same request after 30 seconds gives the same exact response with the same generation date, but with an age header as it was served from cache:

$ curl -i localhost:8000/service/1/valid-for-minute
HTTP/1.1 200 OK
content-type: text/html; charset=utf-8
content-length: 103
cache-control: max-age=60
custom-header: any value
etag: "172ae25df822c3299cf2248694b4ce23"
date: Fri, 11 Sep 2020 03:20:40 GMT
server: envoy
x-envoy-upstream-service-time: 11
age: 30

This response will stay fresh for one minute
Response body generated at: Fri, 11 Sep 2020 03:20:40 GMT

After 1 minute and 1 second:

$ curl -i localhost:8000/service/1/valid-for-minute
HTTP/1.1 200 OK
cache-control: max-age=60
custom-header: any value
etag: "172ae25df822c3299cf2248694b4ce23"
date: Fri, 11 Sep 2020 03:21:41 GMT
server: envoy
x-envoy-upstream-service-time: 8
content-length: 103
content-type: text/html; charset=utf-8

This response will stay fresh for one minute
Response body generated at: Fri, 11 Sep 2020 03:20:40 GMT

The same response was served after being validated with the backend service. You can verify this as the response generation time is the same, but the response date header was updated with the validation response date. Also, no age header.

Every time the response is validated, it stays fresh for another minute. If the response body changes while the cached response is still fresh, the cached response will still be served. The cached response will only be updated when it is no longer fresh.

2. private

$ curl -i localhost:8000/service/1/private
HTTP/1.1 200 OK
content-type: text/html; charset=utf-8
content-length: 117
cache-control: private
etag: "6bd80b59b2722606abf2b8d83ed2126d"
date: Fri, 11 Sep 2020 03:22:28 GMT
server: envoy
x-envoy-upstream-service-time: 7

This is a private response, it will not be cached by Envoy
Response body generated at: Fri, 11 Sep 2020 03:22:28 GMT

No matter how many times you make this request, you will always receive a new response; new date of generation, new date header, and no age header.

3. no-cache

$ curl -i localhost:8000/service/1/no-cache
HTTP/1.1 200 OK
content-type: text/html; charset=utf-8
content-length: 130
cache-control: max-age=0, no-cache
etag: "ce39a53bd6bb8abdb2488a5a375397e4"
date: Fri, 11 Sep 2020 03:23:07 GMT
server: envoy
x-envoy-upstream-service-time: 7

This response can be cached, but it has to be validated on each request
Response body generated at: Fri, 11 Sep 2020 03:23:07 GMT

After a few seconds:

$ curl -i localhost:8000/service/1/no-cache
HTTP/1.1 200 OK
cache-control: max-age=0, no-cache
etag: "ce39a53bd6bb8abdb2488a5a375397e4"
date: Fri, 11 Sep 2020 03:23:12 GMT
server: envoy
x-envoy-upstream-service-time: 7
content-length: 130
content-type: text/html; charset=utf-8

This response can be cached, but it has to be validated on each request
Response body generated at: Fri, 11 Sep 2020 03:23:07 GMT

You will receive a cached response that has the same generation time. However, the date header will always be updated as this response will always be validated first. Also, no age header.

If you change the response body in the yaml file:

$ curl -i localhost:8000/service/1/no-cache
HTTP/1.1 200 OK
content-type: text/html; charset=utf-8
content-length: 133
cache-control: max-age=0, no-cache
etag: "f4768af0ac9f6f54f88169a1f3ecc9f3"
date: Fri, 11 Sep 2020 03:24:10 GMT
server: envoy
x-envoy-upstream-service-time: 7

This response can be cached, but it has to be validated on each request!!!
Response body generated at: Fri, 11 Sep 2020 03:24:10 GMT

You will receive a new response that’s served from the backend service. The new response will be cached for subsequent requests.

You can also add new responses to the yaml file with different cache-control headers and start experimenting!

See also

MDN Web Docs.

Learn more about caching and cache-control on the web.