Java SDK 1.x获取ElastiCache CloudWatch指标返回空值问题

问题 #

使用AWS Java SDK 1.x从CloudWatch获取ElastiCache监控指标时，始终返回0.0，但AWS控制台中该节点的对应指标有有效数据。日志显示GetMetricStatisticsResult的Datapoints为空数组，例如：

Result for scrsquy-001 is {Label: EngineCPUUtilization,Datapoints: []}

相关代码如下：

public class ElastiCacheCWMetricsRetriever {
    private final AmazonCloudWatch cloudWatchClient;

    public ElastiCacheCWMetricsRetriever(AmazonCloudWatch cloudWatchClient) {
        this.cloudWatchClient = cloudWatchClient;
    }


    public Map<String, Map<String, Object>> getElastiCacheMetrics(List<String> nodeNames, 
        String startTimeStr, String endTimeStr) {
        Map<String, Map<String, Object>> elastiCacheMetricsMap = new HashMap<>();

        try {
            SimpleDateFormat dateFormat = new SimpleDateFormat("yyyy-MM-dd'T'HH:mm:ss'Z'");
            Date startTime = dateFormat.parse(startTimeStr);
            Date endTime = dateFormat.parse(endTimeStr);

            for (String nodeName : nodeNames) {
                Map<String, Object> nodeMetrics = new HashMap<>();

                nodeMetrics.put("SwapUsage", getAverageMetricValue("AWS/ElastiCache", 
                    "SwapUsage", "CacheNodeId", nodeName, startTime, endTime));
                Double SwapUsage = getSumMetricValue("AWS/ElastiCache", "SwapUsage", 
                    "CacheNodeId", nodeName, startTime, endTime);
                System.out.println("SwapUsage for " + nodeName + " is " + 
                    String.valueOf(SwapUsage));

                nodeMetrics.put("CurrentConnections", getSumMetricValue("AWS/ElastiCache", 
                    "CurrConnections", "CacheNodeId", nodeName, startTime, endTime));

                nodeMetrics.put("DatabaseMemoryUsagePercentage", 
                getAverageMetricValue("AWS/ElastiCache", "DatabaseMemoryUsagePercentage", 
                    "CacheNodeId", nodeName, startTime, endTime) * 100.0);

                nodeMetrics.put("EngineCPUUtilization", 
                     getAverageMetricValue("AWS/ElastiCache", "EngineCPUUtilization", 
                     "CacheNodeId", nodeName, startTime, endTime) * 100.0);

                Double EngineCPUUtil = getAverageMetricValue("AWS/ElastiCache", 
                    "EngineCPUUtilization", "CacheNodeId", nodeName, startTime, endTime) * 100.0;
                System.out.println("EngineCPUUtilization for " + nodeName + " is " + String.valueOf(EngineCPUUtil));
                elastiCacheMetricsMap.put(nodeName, nodeMetrics);
            }
        } catch (Exception e) {
            System.out.println("Error in Elasticache" + e.getMessage());
        }   
        return elastiCacheMetricsMap;
    }

     private double getAverageMetricValue(String namespace, String metricName, String 
         dimensionName, String dimensionValue, Date startTime, Date endTime) {
         int period = calculatePeriod(startTime, endTime);
         return getMetricValue(namespace, metricName, dimensionName, dimensionValue, startTime, endTime, period, "Average");
     }

    private double getSumMetricValue(String namespace, String metricName, String dimensionName, String dimensionValue, Date startTime, Date endTime) {
        int period = calculatePeriod(startTime, endTime);
        return getMetricValue(namespace, metricName, dimensionName, dimensionValue, startTime, endTime, period, "Sum");
    }

    private double getMetricValue(String namespace, String metricName, String dimensionName, String dimensionValue, Date startTime, Date endTime, int period, String statistic) {
        try {
            GetMetricStatisticsRequest request = new GetMetricStatisticsRequest()
                .withNamespace(namespace)
                .withMetricName(metricName)
                .withDimensions(new Dimension().withName(dimensionName).withValue(dimensionValue))
                .withStartTime(startTime)
                .withEndTime(endTime)
                .withPeriod(period)
                .withStatistics(statistic);

            GetMetricStatisticsResult result = cloudWatchClient.getMetricStatistics(request);
            System.out.println("Result for " + dimensionValue + " is " + result.toString());
            return result.getDatapoints().isEmpty() ? 0 : result.getDatapoints().get(0).getAverage();
        } catch (Exception e) {
            System.out.println("Error in Elasticache" + e.getMessage());
            return 0;
        }
    }


    private int calculatePeriod(Date startTime, Date endTime) {
        long diffInSeconds = (endTime.getTime() - startTime.getTime()) / 1000;
        int periods = (int) Math.ceil(diffInSeconds / 1800);

        return (periods > 0) ? 1800 : 60;
    }
}

排查与解决方案 #

1. 补充完整维度参数 #

ElastiCache的CloudWatch指标需要同时指定**CacheClusterId和CacheNodeId**两个维度，你当前只传入了CacheNodeId，这是导致数据无法匹配的核心原因。控制台能查到数据是因为自动关联了集群ID，但API调用必须明确指定双维度。

修改getMetricValue方法中的维度设置：

.withDimensions(
    new Dimension().withName("CacheClusterId").withValue("你的集群ID"), // 例如从节点名scrsquy-001提取scrsquy
    new Dimension().withName("CacheNodeId").withValue(dimensionValue)
)

2. 修正统计值与指标的匹配关系 #

部分指标的统计值选择错误，导致返回空数据：

• SwapUsage、CurrConnections、DatabaseMemoryUsagePercentage、EngineCPUUtilization均需使用Average统计值，Sum对这些指标无意义，会返回空。
• 移除代码中对SwapUsage、CurrConnections的Sum统计调用，统一使用Average。

3. 调整周期计算逻辑 #

当前周期计算可能生成CloudWatch不支持的数值，或超出指标的聚合范围。CloudWatch支持的周期为60、300、3600等秒数，且需能整除时间范围的总秒数：

private int calculatePeriod(Date startTime, Date endTime) {
    long diffInSeconds = (endTime.getTime() - startTime.getTime()) / 1000;
    // 优先使用300秒（5分钟），时间范围小于5分钟则用60秒
    if (diffInSeconds <= 60) {
        return 60;
    } else if (diffInSeconds <= 300) {
        return 60;
    } else {
        return 300;
    }
}

4. 修正时区解析问题 #

SimpleDateFormat默认使用本地时区，而CloudWatch时间为UTC，会导致时间范围偏差。解析时指定UTC时区：

SimpleDateFormat dateFormat = new SimpleDateFormat("yyyy-MM-dd'T'HH:mm:ss'Z'");
dateFormat.setTimeZone(TimeZone.getTimeZone("UTC"));

5. 修复统计值获取逻辑 #

当前代码无论请求的是Sum还是Average，都调用getAverage()获取值，会导致数据错误。修改getMetricValue方法：

return result.getDatapoints().isEmpty() ? 0 : 
    "Sum".equals(statistic) ? result.getDatapoints().get(0).getSum() : result.getDatapoints().get(0).getAverage();

内容的提问来源于stack exchange，提问作者davidb