Presume an index contains docs with a single text field. The values of that field can be arbitrarily huge docs. There can easily come a point where doing a simple search against that index and requesting an appropriate size of docs in a single response can OOM the node.
Could there be a circuit breaker that prevents the node from OOM?
Steps to reproduce:
a dataset which the message field in some documents have the length larger than 700,000. (please contact @wwang500 for datasets)
Description
Presume an index contains docs with a single text field. The values of that field can be arbitrarily huge docs. There can easily come a point where doing a simple search against that index and requesting an appropriate
size
of docs in a single response can OOM the node.Could there be a circuit breaker that prevents the node from OOM?
Steps to reproduce:
Observed: Node will turn unhealthy and restart due to OOM.