News KrakenD CE v2.8 released with improved Lua and OpenTelemetry

You are viewing a previous version of KrakenD Community Edition (v2.0), go to the latest version

Document updated on May 2, 2021

Caching backend responses

Sometimes you might want to reuse a previous response of a backend instead of asking for the same information over the network again. In this cases, it is possible to enable in-memory caching for the desired backend responses.

This caching technique applies to traffic between KrakenD and your microservices endpoints only and is not a caching system for the end-user endpoints. To enable the cache, you only need to add in the configuration file the qos/httpcache middleware.

When enabled all connections for the configured backend are cached in-memory. The cache content is based on the response for the final URL sent to the backend (the url_pattern plus any additional parameters). The response is stored for the time the Cache-Control has defined and there is no way to purge it externally.

Performance notice!
This option can increase the load and memory consumption heavily as KrakenD needs to keep in memory all the returned data during the expiration period. Only the backend has control over this. Use it wisely and monitor its consumption!

If you enable this module, you are required to be very aware of the response sizes, caching times and the hit-rate of the calls.

Enable the caching of the backend services in the backend section of your krakend.json with the middleware:

{
  "extra_config": {
    "qos/http-cache": {}
  }
}

The middleware does not require additional configuration other than its simple inclusion, although the shared attribute can be passed.

See an example (with shared cache):

With shared cache:

{
    "endpoint": "/cached",
    "backend": [{
      "url_pattern": "/",
      "host": ["http://my-service.tld"],
      "extra_config": {
        "qos/http-cache": {
            "shared": true
        }
      }
    }]
}

The shared cache makes that different backend definitions with this flag enabled can reuse the cache. When the shared flag is missing or set to false, the backend uses its own cache private context.

Scarf

Unresolved issues?

The documentation is only a piece of the help you can get! Whether you are looking for Open Source or Enterprise support, see more support channels that can help you.

See all support channels