How to check cache hits/misses with jsonapi-rails

Question

Currently using jsonapi-rails to render serialized objects while utilizing some isolated cache.

JSON.generate JSONAPI::Rails::SuccessRenderer.new.render(
  all_data,
    {
      include: %w(attr_1 attr_2),
      meta: {
        total: all_data.size
      },
      expose: {
        url_helpers: ::Rails.application.routes.url_helpers,
        language: language,
      },
      cache: SerializedObjectsCache.instance
    },
  controller
)

# SerializedObjectsCache
def self.instance
    @instance ||=
      begin
        cache = ActiveSupport::Cache::RedisCacheStore.new(
          redis: Redis.current,
          compress: COMPRESS,
          namespace: 'serialized_objects',
          expires_in: EXPIRES_IN
        )
        new(cache)
      end
  end

What I'm trying to figure out is how often we're missing/hitting the cache. Each serialized object has a jsonapi_cache_key method to return a unique key for each object but I'm unsure of how to see the cache hit/miss rate.

Answer 1

Just a suggestion, I'm not 100% sure I understand your goals here.

You code uses ActiveSupport::Cache::RedisCacheStore, which is instrumented. To be more precise, it publishes cache_read_multi.active_support notification when one calls read_multi (fetch_multi that renderer calls does that under the hood). The notification payload includes hits data (quite dirty rudimentary example):

pry(main)> c = ActiveSupport::Cache::RedisCacheStore.new
pry(main)> ActiveSupport::Notifications.subscribe("cache_read_multi.active_support") { |_, _, _, _, p| puts p.inspect }
pry(main)> c.fetch_multi(:foo) { "default foo" }
"cache_read_multi.active_support"
["cache_read_multi.active_support", "2022-02-03T13:34:12.522+01:00", "2022-02-03T13:34:12.531+01:00", "c4372c000b2f8d1dc867", {:key=>[:foo], :namespace=>nil, :compress=>true, :compress_threshold=>1024, :expires_in=>nil, :race_condition_ttl=>nil, :hits=>[:foo], :super_operation=>:fetch_multi}]
{:key=>[:foo], :namespace=>nil, :compress=>true, :compress_threshold=>1024, :expires_in=>nil, :race_condition_ttl=>nil, :hits=>[:foo], :super_operation=>:fetch_multi}
=> {:foo=>"default foo"}

So, having hits at hand you should be able to calculate the miss/hit rate - but some additional boilerplate is required to aggregate the data for a certain period.

The biggest gotcha here is that adding this additional machinery to accumulate hits stats might kill the performance gains from the caching itself.