Proxy Caching Advanced

Did you know that you can try this plugin without talking to anyone with a free trial of Kong Konnect? Get started in under 5 minutes.

This plugin provides a reverse proxy cache implementation for Kong. It caches response entities based on configurable response code and content type, as well as request method. It can cache per-Consumer or per-API. Cache entities are stored for a configurable period of time, after which subsequent requests to the same resource will re-fetch and re-store the resource. Cache entities can also be forcefully purged via the Admin API prior to their expiration time.

Warning: The content_type parameter requires an exact match. For example, if your upstream expects application/json; charset=utf-8 and the config.content_type value is only application/json (a partial match), then the proxy cache is bypassed.

Strategies

kong-plugin-enterprise-proxy-cache is designed to support storing proxy cache data in different backend formats. Currently, the following strategies are provided:

• memory: A lua_shared_dict. Note that the default dictionary, kong_db_cache, is also used by other plugins and elements of Kong to store unrelated database cache entities. Using this dictionary is an easy way to bootstrap the proxy-cache-advanced plugin, but it is not recommended for large-scale installations as significant usage will put pressure on other facets of Kong’s database caching operations. It is recommended to define a separate lua_shared_dict via a custom Nginx template at this time.
• redis: Supports Redis and Redis Sentinel deployments.

Cache Key

Kong keys each cache elements based on the request method, the full client request (e.g., the request path and query parameters), and the UUID of either the API or Consumer associated with the request. This also implies that caches are distinct between APIs and/or Consumers. Currently the cache key format is hard-coded and cannot be adjusted. Internally, cache keys are represented as a hexadecimal-encoded MD5 sum of the concatenation of the constituent parts calculated as follows:

key = md5(UUID | method | request)

Where method is defined via the OpenResty ngx.req.get_method() call, and request is defined via the Nginx $request variable. Kong will return the cache key associated with a given request as the X-Cache-Key response header. It is also possible to precalculate the cache key for a given request as noted above.

Cache Control

When the cache_control configuration option is enabled, Kong will respect request and response Cache-Control headers as defined by RFC7234, with a few exceptions:

• Cache revalidation is not yet supported, and so directives such as proxy-revalidate are ignored.
• Similarly, the behavior of no-cache is simplified to exclude the entity from being cached entirely.
• Secondary key calculation via Vary is not yet supported.

Cache Status

Kong identifies the status of the request’s proxy cache behavior via the X-Cache-Status header. There are several possible values for this header:

• Miss: The request could be satisfied in cache, but an entry for the resource was not found in cache, and the request was proxied upstream.
• Hit: The request was satisfied and served from cache.
• Refresh: The resource was found in cache, but could not satisfy the request, due to Cache-Control behaviors or reaching its hard-coded cache_ttl threshold.
• Bypass: The request could not be satisfied from cache based on plugin configuration.

Storage TTL

Kong can store resource entities in the storage engine longer than the prescribed cache_ttl or Cache-Control values indicate. This allows Kong to maintain a cached copy of a resource past its expiration, which in turn allows clients capable of using max-age and max-stale headers to request stale copies of data if necessary.

Upstream Outages

Due to an implementation in Kong’s core request processing model, at this point, the proxy-cache-advanced plugin cannot be used to serve stale cache data when an upstream is unreachable. To equip Kong to serve cache data in place of returning an error when an upstream is unreachable, we recommend defining a very large storage_ttl (on the order of hours or days) in order to keep stale data in the cache. In the event of an upstream outage, stale data can be considered fresh by increasing the cache_ttl plugin configuration value. By doing so, data that would have previously been considered stale is now served to the client, before Kong attempts to connect to a failed upstream service.

Admin API

This plugin provides several endpoints to managed cache entities. These endpoints are assigned to the proxy-cache-advanced RBAC resource.

The following endpoints are provided on the Admin API to examine and purge cache entities:

Retrieve a Cache Entity

Two separate endpoints are available: one to look up a known plugin instance, and another that searches all proxy-cache-advanced plugins data stores for the given cache key. Both endpoints have the same return value.

Endpoint

Endpoint

Response

If the cache entity exists

HTTP 200 OK

If the entity with the given key does not exist

HTTP 400 Not Found

Delete a Cache Entity

Two separate endpoints are available: one to look up a known plugin instance, and another that searches all proxy-cache-advanced plugins data stores for the given cache key. Both endpoints have the same return value.

Endpoint

Endpoint

Response

If the cache entity exists:

HTTP 204 No Content

If the entity with the given key does not exist:

HTTP 400 Not Found

Purge All Cache Entities

Endpoint

Response

HTTP 204 No Content

Note that this endpoint purges all cache entities across all proxy-cache-advanced plugins.