Large memory footprint: Sentinel MetricNode

[zcwdlove]

Describe what happened (or what feature you want)

Large memory footprint detected (objects of Sentinel)

线上发布应用后查看堆栈信息发现sentinel相关对象占用太高, 项目中除了测试用例几乎没有埋点信息，请问如何优化以减少内存占用？

详情见图 (heap objects):

Tell us your environment

Sentinel 1.5.2

[sczyh30]

This might be caused by inappropriate metric item generating in StatisticNode:

Sentinel/sentinel-core/src/main/java/com/alibaba/csp/sentinel/node/StatisticNode.java

Line 121 in d45ceb0

List<MetricNode> nodesOfEverySecond = rollingCounterInMinute.details();

Every second Sentinel will aggregate metrics of every resource and write to the metric log, but when aggregating the metric of last second, it actually generates metric items in the past minute (i.e. 60s), which is not necessary. This could be improved, but the handling logic of fetchTime should be considered.

Discussions and contributions are welcomed!

[sczyh30]

Since 1.7.0 we've provided a useful method rawMetricsInMin in Node interface, which could help us to improve the code:

Sentinel/sentinel-core/src/main/java/com/alibaba/csp/sentinel/node/StatisticNode.java

Lines 136 to 138 in d45ceb0

	public List<MetricNode> rawMetricsInMin(Predicate<Long> timePredicate) {
	return rollingCounterInMinute.detailsOnCondition(timePredicate);
	}

The legacy method metrics could be deprecated.

[xzy-max1]

Has this problem been solved?