Negative cache miss metric

Hi,

I’m registering sme Hazelcast cache metrics using

HazelcastCacheMetrics.monitor(meterRegistry, hazelcastInstance.getMap(Constants.METADATA_DEFINITION_CACHE));

And then using then like this:

    @CachePut(value = Constants.METADATA_DEFINITION_CACHE, key = "#metadataDefinitionId", unless = "#result == null")
    public MetadataDefinition putMetadataDefinition(long metadataDefinitionId) {
        return getMetadataDefinitionMetadataInternal(metadataDefinitionId);
    }

    @Cacheable(value = Constants.METADATA_DEFINITION_CACHE, key = "#metadataDefinitionId", unless = "#result == null")
    public MetadataDefinition getMetadataDefinition(long metadataDefinitionId) {
        return getMetadataDefinitionMetadataInternal(metadataDefinitionId);
    }

    @CacheEvict(value = Constants.METADATA_DEFINITION_CACHE, key = "#metadataDefinitionId")
    public void removeFromCache(long metadataDefinitionId) {
    }

but somehow I’m getting negative miss values.

# HELP cache_gets_total the number of times cache lookup methods have returned an uncached (newly loaded) value, or null
# TYPE cache_gets_total counter
cache_gets_total{cache="metadataDefinitionCache",result="miss",} -156909.0
cache_gets_total{cache="metadataDefinitionCache",result="hit",} 314086.0

Any ideas why?

Thanks

Nick

[mweirauch]

Just as a hint, when somebody touches the file, the class Javadoc mentions “EhCache”. Likely a C&P leftover.

[jkschneider]

@nickcodefresh Without your screenshots, this would have been very hard to track down. In the end, I concluded that it is impossible to calculate miss count against Hazelcast caches 😦

I’ve removed the metric from the Hazelcast implementation, and added another FunctionCounter, cache.partition.gets.

See this for the detail on why miss count is incalculable on Hazelcast.

If it is really important for you to know miss count, you can create a wrapper IMap type that adds to a counter every time get is about to return null.