之前已经监控了很多metrics,现在我们需要监控metrics自身的一些指标,例如我们比较关注的是其每个指标的基数。基数太大会导致客户端和服务端内存溢出问题。
相关代码如下
import io.micrometer.core.instrument.Gauge;
import io.micrometer.core.instrument.Meter;
import io.micrometer.core.instrument.composite.CompositeMeterRegistry;
import org.springframework.scheduling.annotation.Scheduled;
import org.springframework.stereotype.Component;
import java.util.List;
import java.util.Map;
import java.util.concurrent.ConcurrentHashMap;
import java.util.stream.Collectors;
@Component
public class MetricsCardinalityScheduledTask {
private final Map<String, Integer> meterCardinalityMap = new ConcurrentHashMap<>();
private final CompositeMeterRegistry compositeMeterRegistry;
public MetricsCardinalityScheduledTask(CompositeMeterRegistry compositeMeterRegistry) {
this.compositeMeterRegistry = compositeMeterRegistry;
}
@Scheduled(fixedDelayString = "#{@lakerProperties.scheduleTaskDelayMilliseconds}")
public void emitCardinalityForEachMeasurement() {
// 清空
meterCardinalityMap.clear();
// 剔除自身的meter。
List<Meter> registryMeters = compositeMeterRegistry.getMeters()
.stream()
.filter(meter -> !meter.getId().getName().equals("measurement.cardinality"))
.collect(Collectors.toList());
// 转换所有的Meters
registryMeters
.forEach(meter -> meterCardinalityMap.merge(meter.getId().getName(), 1, Integer::sum));
for (String measurement : meterCardinalityMap.keySet()) {
Gauge.builder("measurement.cardinality",
() -> (long) meterCardinalityMap.get(measurement))
.tags("measurement", measurement)
.register(compositeMeterRegistry);
}
}
}
为什么上面的代码就可以统计基数呢?
为此我们针对Meter解释下。
Meter是以Id为唯一key,例如有2个指标mem,cpu
mem
cpu
则Meter有4个.指标2个为mem,cpu
| name | id |
|---|---|
| mem | mem:10.0.0.1 |
| mem | mem:10.0.0.2 |
| cpu | cpu:10.0.0.1 |
| cpu | cpu:10.0.0.2 |