Es restore [parent] data too large

The total data size of es is less than 1G. Perform a regular restore. After running for a period of time, throw the [parent] data too large. The node executing the restore has a high xmx and a high CPU utilization rate_ After analyzing nodes/stat/breaker, it was found that only parent estimated_ size_ in_ The usage of bytes exceeds the limit. Other examples such as requests and accounting are very small.

ES 7.10.0

You are running a very old version, so I would recommend that you upgrade to at least version 7.17 as soon as possible.

What is the full output of the cluster stats API?

Is this a known problem? It appears that heap memory has not been freed.

MEMORY_ MX_ BEAN.getHeapMemoryUsage().getUsed();

Please answer this question.

The picture shows some validation I did in the test environment; parent estimated_ Size takes up a lot of memory.

Set indices.breaker.total.use_real_memory=false, parent estimated_size only occupies about 200 kb, and is back to normal.

If i set indices.breaker.total.use_real_memory=false, is there anything i need to pay attention to?

Please do not show pictures of text. Copy the response and paste it here instead.

"breaker": {
	"request": {……},
	"fielddata": {……},
	"in_flight_requests": {……},
	"model_inference": {……},
	"accounting": {……},
	"parent": {
		"limit_size_in_bytes": 2040109465,
		"limit_size": "1.8gb",
		"estimated_size_in_bytes": 845819696,
		"estimated_size": "806.6mb",
		"overhead": 1,
		"tripped": 0
	}
}

(request、fielddata、in_flight_requests、model_inference、accounting) memory usages very low; but parent memory usage very high,until data too large.

There is a lot of useful information within the response that can help point to the issue, so please post the full response. What you posted is not at all what I am looking for.

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.