Hit rate tracking for Magento Page Cache.

All configuration is done through an XML file in the /app/etc directory. This is because module configuration is not loaded in the case of a full hit. mpchadwick_pagecachehitrate.xml.template is included with example settings. A few notes...

• Mpchadwick_PageCacheHitRate_Model_Processor is added as a <request_processor>.
  • Note: request_processors are loaded alphabetically based on the file name in /app/etc. It is important that Mpchadwick_PageCacheHitRate_Model_Processor be the final request_processor to know for sure if this is a full hit. By default the Enterprise_PageCache_Model_Processor is defined in enterprise.xml and will be loaded first, however if your are using something else to process the result of Enterprise_PageCache_Model_Processor you may need to change the file name in /app/etc.
• <tracker>s can be configured in your <full_page_cache> configuration under the <mpchadwick_pagecachehitrate> node. If no trackers are specified, hit rate will not be tracked.
• There is a <track_container_misses> setting which can be used to track individual container misses in the case of partial page hits.
• A <metadata_source> can be configured under the <full_page_cache> configuration in the case where Enterprise_PageCache_Model_Processor is not the correct class to provide metadata. This can happen if the <ee> request processor is changed, notably with Elastera_EnterprisePageCache.

The following trackers are available...

• Mpchadwick_PageCacheHitRate_Model_Tracker_File A log file will be used for for tracking hit rate. A new file will be created each day.
• Mpchadwick_PageCacheHitRate_Model_Tracker_NewRelic New Relic custom events will be used for tracking hit rate.
• Mpchadwick_PageCacheHitRate_Model_Tracker_Redis Hit rate will be tracked in Redis. The intention is to scrape this data periodically and store it in a tool such as InfluxDb or Prometheus.

You can easily create your own tracker if you'd prefer a different means of tracking. Simply implement the Mpchadwick_PageCacheHitRate_Model_TrackerInterface interface and configure your class as the <tracker> in an xml file in /app/etc.

In addition to a full hit and a full miss, Enterprise_PageCache can also have partial hits. This happens if a cached response is found, but it has containers that need additional processing. The type metric is thus tracked as either a hit, a miss, or a partial.

In addition to type, the following dimensions are tracked...

• route,
• url
• ip
• hostname
• customerGroup

For example, you may want to see your cache hit rate for a single route such as /catalog/product/view or know your cache hit rate for a single IP address if your are seeing crawling / bot activity creating performance impact on your site.

Here are some example entries using the Mpchadwick_PageCacheHitRate_Model_Tracker_File tracker...

2016-06-02T02:28:04+00:00 DEBUG (7): {"url":"http:\/\/magento-1_14_1_0.dev\/men\/shirts.html","ip":"172.16.9.1","type":"miss","route":"catalog\/category\/view"}
2016-06-02T02:28:09+00:00 DEBUG (7): {"url":"http:\/\/magento-1_14_1_0.dev\/men\/shirts.html","ip":"172.16.9.1","type":"partial","route":"catalog\/category\/view"}
2016-06-02T02:28:11+00:00 DEBUG (7): {"url":"http:\/\/magento-1_14_1_0.dev\/men\/shirts.html","ip":"172.16.9.1","type":"hit","route":"catalog\/category\/view"}

Container misses will be recorded to a separate file (if enabled). The entries will look like this...

2016-06-02T03:18:14+00:00 DEBUG (7): {"url":"http:\/\/magento-1_14_1_0.dev\/women\/women-new-arrivals.html","ip":"172.16.9.1","route":"catalog\/category\/view","container":"Enterprise_PageCache_Model_Container_Catalognavigation"}

Also note that you can configure multiple trackers. You may want to store data in both New Relic and Redis, for example.

By default, highly cardinal dimensions are tracked such as IP address and URL. For the Mpchadwick_PageCacheHitRate_Model_Tracker_File and Mpchadwick_PageCacheHitRate_Model_Tracker_NewRelic trackers this is probably fine. However if you are using Mpchadwick_PageCacheHitRate_Model_Tracker_Redis and sending the data to InfluxDb or Prometheus you'll probably want to decrease the cardinality of the data.

You can specify dimensions to strip on a per tracker-basis. E.g...

<config>
    <global>
        <full_page_cache>
            <mpchadwick_pagecachehitrate>
                <trackers>
                    <redis>
                        <class>Mpchadwick_PageCacheHitRate_Model_Tracker_Redis</class>
                        <server>127.0.0.1</server>
                        <port>6379</port>
                        <database>1</database>
                        <strip>
                            <ip />
                            <url />
                        </strip>
                    </redis>
                </trackers>
            </mpchadwick_pagecachehitrate>
        </full_page_cache>
    </global>
</config>