Elasticsearch - Different between indices.fielddata.cache.size and indices.fielddata.breaker.limit

Question

One of the biggest causes of instability in Elasticsearch is fielddata: field values have to be loaded into memory to make aggregations, sorting and scripting perform as fast as they do.

As description above on Elasticsearch page, large fielddata always causes Elasticsearch out of memory(OOM). Thus we can set indices.fielddata.cache.size and indices.fielddata.breaker.limit to prevent OOM. What is the different between those two settings? Are they have any relation?

For example, My Elasticsearch JVM has 2g total memory. If I set indices.fielddata.cache.size to 1g but indices.fielddata.breaker.limit set to 60% (Which means 1.2g). The fielddata allowed to load to memory is over the fielddata cache size. Is it will causes any error? (Reference Fielddata)

Thank you.

Answer 1

After study for a long time, I found some answer.

When you set indices.fielddata.cache.size to 1g. It means how many field cache size elasticsearch can use to handle query request. But when you set indices.fielddata.breaker.limit to 60% (means 1.2g), If the query data is larger than this size, elasticsearch will reject this query request and causes an exception.

So, if query data is smaller than 1.2g but larger than 1g, elassticsearch will accept this query request. After reach the indices.fielddata.cache.size, the old data will be flush and release the memory for new data.

Answer 2

Their difference is, I quote

Fielddata size is checked after the data is loaded. What happens if a query arrives that tries to load more into fielddata than available memory? The answer is ugly: you would get an OutOfMemoryException.

Elasticsearch includes a fielddata circuit breaker that is designed to deal with this situation. The circuit breaker estimates the memory requirements of a query by introspecting the fields involved (their type, cardinality, size, and so forth). It then checks to see whether loading the required fielddata would push the total fielddata size over the configured percentage of the heap.

If the estimated query size is larger than the limit, the circuit breaker is tripped and the query will be aborted and return an exception. This happens before data is loaded, which means that you won’t hit an OutOfMemoryException.

from Limiting Memory Usage.