java - CloudWatch Alarm 进入针对单个数据点的 ALARM

Question

如果平均性能超过 5 毫秒，我将尝试创建 cloudwatch 警报。我已配置 Amazon CloudWatch 警报来检查平均值。但是，如果只有一个数据点超过阈值，警报就会进入“警报”状态。

public static void main(String[] args) {
   AWSExample aws = new AWSExample();
   aws.testMethod();
}

这是测试方法。

public void testMethod() {
        Instant start = Instant.now();
        try {
            try {
            long myValue = (long) ((Math.random())*10000);
            if(myValue>8000){
                myValue = myValue - 3000;
            }
            Thread.sleep(myValue);
            } catch(InterruptedException ex) {
                Thread.currentThread().interrupt();
            }
        } catch (Throwable t) {
            t.printStackTrace();
        } finally {
            Instant end = Instant.now();
            Duration nano = Duration.between(start, end);
            long endTime = nano.toMillis();
            createMetricData(endTime);
            createAnAlarm();
        }
    }

创建指标数据的方法

public void createMetricData(Long metricValue) {
        final AmazonCloudWatch cw = getAmazonCloudWatch();
        Dimension dimension = new Dimension().withName("UNIQUE_METHOD").withValue("testMethod");
        MetricDatum datum = new MetricDatum()
                .withMetricName("Method Execution Performance")
                .withUnit(StandardUnit.Milliseconds).withValue(metricValue.doubleValue())
                .withDimensions(dimension)
                .withTimestamp(new Date());
        PutMetricDataRequest metricDataRequest = new PutMetricDataRequest()
                .withNamespace("METHOD/TRAFFIC").withMetricData(datum);
        PutMetricDataResult response = cw.putMetricData(metricDataRequest);
        System.out.println(response);
        System.out.printf("Successfully put data point %f", metricValue.doubleValue());
    }

这是创建警报的方法。

private void createAnAlarm(){
        final AmazonCloudWatch cw = getAmazonCloudWatch();
        PutMetricAlarmRequest putMetricAlarmRequest = new PutMetricAlarmRequest()
        .withPeriod(120)// The period, in seconds, over which the specified statistic is applied. Valid values are 10, 30, and any multiple of 60.
        .withMetricName("Method Execution Performance")// The name for // the metric // associated // with the // alarm.
        .withNamespace("METHOD/TRAFFIC")// The namespace for the metric // associated with the alarm. // https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/aws-namespaces.html
        .withAlarmName("aws-Method-Performance")// The name for the alarm. // This name must be unique // within the AWS account.
        .withEvaluationPeriods(1)// The number of periods over which // data is compared to the specified // threshold. An alarm's total // current evaluation period can be // no longer than one day, so this
                                    // number multiplied by period cannot be more than 86,400 seconds.
        .withActionsEnabled(true)// Indicates whether actions should be executed during any changes to
                                    // the alarm state.
        .withStatistic(Statistic.Average)// The statistic for the metric associated with the alarm, other than percentile
        .withThreshold(5.0)// The value against which the specified statistic is compared.
        .withComparisonOperator(ComparisonOperator.GreaterThanThreshold)
        .withAlarmDescription("Alarm when method execution time exceeds 5 milliseconds")
        .withAlarmActions("arn:aws:sns:eu-west-1:***********")//The actions to
        //execute when this alarm transitions to the ALARM state from
        // any other state. Each action is specified as an Amazon
        // Resource Name (ARN).
        .withUnit(StandardUnit.Milliseconds)// The unit of measure for
                                                // the statistic
        .withDimensions(new Dimension().withName("UNIQUE_METHOD").withValue("testMethod"));
        PutMetricAlarmResult result = cw.putMetricAlarm(putMetricAlarmRequest);
        System.out.println(result);
    }

我希望该方法的平均性能超过 5 毫秒。这里有什么问题以及如何解决它？

Answer 1

所以问题是，如果只有一个数据点突破阈值，为什么警报会进入“警报”状态？那是因为您在警报创建中有这一行:

.withEvaluationPeriods(1)

此外，您每次发布数据点时都会调用 createAnAlarm();。无需这样做，您只需创建一次警报，它就会持续监控您的指标。

正如评论中所讨论的，触发此警报的实际原因是阈值设置为 5 毫秒，但该方法的预期执行时间在秒范围内。在这种情况下设置的正确阈值是 5 秒:

.withThreshold(5000.0)