[GEODE-751] RegionMXBean shouldn't rely on Eviction Algorithm for getEntrySize

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 1.0.0-incubating
• Component/s: jmx, statistics
• Assignee: Jens Deppe
• Reporter: Jens Deppe
• Created: 07/Jan/16 20:51
• Updated: 20/Aug/19 23:20

Description

We have the following in the javadoc for method getEntrySize on interface RegionMXBean:

Returns the aggregate entry size (in bytes) of all entries. This will provide a correct value only if the eviction algorithm has been set to EvictionAlgorithm.LRU_MEMORY. For all partition regions it will show entry size in bytes. It will also include size of all the secondary entries in the data store. So while referring to size one should take redundancy into account.

The region memory consumption and the eviction algorithm are two separate concepts, we should not obligate customers to use a custom eviction algorithm to report the correct memory consumption for a region.

We rely on this attribute to show information on PULSE, so neither the member memory usage nor cluster memory usage are accurate if the eviction algorithm is not configured as EvictionAlgorithm.LRU_MEMORY.

To reproduce, start up a cluster with a simple replicated region and insert some data. You can check afterwards (from JConsole) that the attribute "EntrySize" for the replicated region is set as "-1".