探索Windows Azure 监控和自动伸缩系列2

上一篇博文介绍了如何连接Windows Azure: http://www.cnblogs.com/teld/p/5113063.html

本篇我们继续上次的示例代码，获取虚拟机的监控定义和监控数据。

有人会问，Azure Portal上已经有了监控数据，通过代码获取有意思吗？我们计划基于性能计数器的监控数据来实现应用的自动伸缩，因此可以获取到监控指标定义和监控数据应该是第一步。

在Azure的管理Portal中我们可以看到虚拟机的监控数据，目前，提供的主要有以下监控指标：

CPU Percentage；Disk Read; Disk Write; Network in；NetWork Out。

Azure中监控的Nuget主要是这个：Microsoft Azure Management Libraries

核心的几个namespace有：

我们本篇用的是Metric这个命名空间，核心类MetricClient：

namespace AzureTest
{
    using Microsoft.WindowsAzure;
    using Microsoft.WindowsAzure.Management.Monitoring.Metrics;
    using Microsoft.WindowsAzure.Management.Monitoring.Metrics.Models;
    using Microsoft.WindowsAzure.Management.Monitoring.Utilities;

    /// <summary>
    /// 监控客户端
    /// </summary>
    class MonitorClient
    {
        private SubscriptionCloudCredentials credentials;

        public MonitorClient(SubscriptionCloudCredentials credentials)
        {
            this.credentials = credentials;
        }

        /// <summary>
        /// 获取所有的监控指标
        /// </summary>
        public void GetMetricDefinitions()
        {
            var metricsClient = new MetricsClient(credentials);
            // Build the resource ID string.
            var resourceId = ResourceIdBuilder.BuildVirtualMachineResourceId("cloudServiceName", "deploymentName");
            Console.WriteLine("Resource Id: {0}", resourceId);

            //Get the metric definitions.
            var metricListResponse=
                metricsClient.MetricDefinitions.List(resourceId, null, null);34 
            MetricDefinitionCollection metricDefinitions = metricListResponse.MetricDefinitionCollection;           
            // Display the metric definitions.
            int count = 0;
            foreach (MetricDefinition metricDefinition in metricDefinitions.Value)
            {
                Console.WriteLine("MetricDefinitio: " + count++);
                Console.WriteLine("Display Name: " + metricDefinition.DisplayName);
                Console.WriteLine("Metric Name: " + metricDefinition.Name);                
                Console.WriteLine("Metric Namespace: " + metricDefinition.Namespace);                
                Console.WriteLine("Is Altertable: " + metricDefinition.IsAlertable);
                Console.WriteLine("Min. Altertable Time Window: " + metricDefinition.MinimumAlertableTimeWindow);
                Console.WriteLine();
            }
        }
}
}

使用上一篇我们的Azure 凭据验证器，获取一个令牌凭据TokenCloudCredentials，然后构造一个MonitorClient，获取指定虚拟机的监控数据。

static void Main(string[] args)
{            
      var credential = Authorizator.GetCredentials();
      var client = new MonitorClient(credential);
      client.GetMetricDefinitions();
      Console.ReadLine();
}

第一块代码中：

var resourceId = ResourceIdBuilder.BuildVirtualMachineResourceId("cloudServiceName", "deploymentName");

这个地方通ResourceIDBuilder获取虚拟机的资源ID，对应的参数分别为：cloudServiceName和deploymentName，第一个是虚拟机使用的云服务名称，第二个是虚拟机名称即可。

Run...

出错了：

{"ForbiddenError: The server failed to authenticate the request. Verify that the certificate is valid and is associated with this subscription."}

一番Google后未果，咋整，再分析下错误信息：

可以看到，请求的Uri：

{https://management.core.windows.net/37*****-5107-*****-*******6/services/monitoring/metricdefinitions/query?&resourceId=%2Fhostedservices%2Fteldptapp%2Fdeployments%2Fteldptapp}

请求又跑到Azure Global那去了。

这个错误困扰了好久，还在StackOverflow上发了英文咨询贴，不知道洋人们如何回答了。在此多谢鞠强老大的指导，想办法将请求的Uri定位到中国区的Azure。

重新分析了代码，找到了Monitor的构造函数中，可以指定Uri，将中国区Azure的Uri指定一下：https://management.core.chinacloudapi.cn

MetricsClient metricsClient = new MetricsClient(credentials, new Uri("https://management.core.chinacloudapi.cn/"));

测试通过，ok。

获取到了监控指标定义，接下来我们获取监控数据：

namespace AzureTest
{
    using Microsoft.WindowsAzure;
    using Microsoft.WindowsAzure.Management.Monitoring.Metrics;
    using Microsoft.WindowsAzure.Management.Monitoring.Metrics.Models;
    using Microsoft.WindowsAzure.Management.Monitoring.Utilities;
    /// <summary>
    /// 监控客户端
    /// </summary>
    class MonitorClient
    {
        private SubscriptionCloudCredentials credentials;

        public MonitorClient(SubscriptionCloudCredentials credentials)
        {
            this.credentials = credentials;
        }

        /// <summary>
        /// 获取所有的监控指标数据
        /// </summary>
        public void GetMetricData()
        {
            var metricsClient = new MetricsClient(credentials, new Uri("https://management.core.chinacloudapi.cn/"));
            
            // Build the resource ID string.
            var resourceId = ResourceIdBuilder.BuildVirtualMachineResourceId("cloudServiceName", "deploymentName");
            Console.WriteLine("Resource Id: {0}", resourceId);

            //Get the metric definitions.
            var metricListResponse = metricsClient.MetricDefinitions.List(resourceId, null, null);
            MetricDefinitionCollection metricDefinitions = metricListResponse.MetricDefinitionCollection;
            
            var metricNamespace = "";
            var metricNames = new List<string>();
            // Display the metric definitions.
            int count = 0;
            foreach (MetricDefinition metricDefinition in metricDefinitions.Value)
            {
                Console.WriteLine("MetricDefinitio: " + count++);
                Console.WriteLine("Display Name: " + metricDefinition.DisplayName);
                Console.WriteLine("Metric Name: " + metricDefinition.Name);
                if (!metricNames.Contains(metricDefinition.Name))
                    metricNames.Add(metricDefinition.Name);
                Console.WriteLine("Metric Namespace: " + metricDefinition.Namespace);
                metricNamespace = metricDefinition.Namespace;
                Console.WriteLine("Is Altertable: " + metricDefinition.IsAlertable);
                Console.WriteLine("Min. Altertable Time Window: " + metricDefinition.MinimumAlertableTimeWindow);
                Console.WriteLine();
            }

            // timeGrain must be 5, 60 or 720 minutes.
            TimeSpan timeGrain = TimeSpan.FromMinutes(5);
            DateTime startTime = DateTime.UtcNow.AddHours(-1);
            DateTime endTime = DateTime.UtcNow;

            MetricValueListResponse response = metricsClient.MetricValues.List(resourceId, metricNames, metricNamespace, timeGrain, startTime, endTime);

            foreach (MetricValueSet value in response.MetricValueSetCollection.Value)
            {
                String valueName = value.Name;
                Console.WriteLine("MetricValue:{0}", valueName);
                foreach (MetricValue metricValue in value.MetricValues)
                {
                    Console.WriteLine("Maximum:{0}{1}", metricValue.Maximum, value.Unit);
                    Console.WriteLine("Average:{0}{1}", metricValue.Average, value.Unit);
                    Console.WriteLine("Minimum:{0}{1}", metricValue.Minimum, value.Unit);
                }
            }
        }
    }
}

Run...

程序在metricsClient.MetricValues.List(resourceId, metricNames, metricNamespace, timeGrain, startTime, endTime);

出错了：
Additional information: <string xmlns="http://schemas.microsoft.com/2003/10/Serialization/">{"Code":"InvalidRequest","Message":"Could not retrieve metrics."}</string>

为啥不能获取监控指标呢？

这个错误，Google一番依旧未果，咨询了微软的技术工程师，给了如下指导，resouceID必须执行RoleName：

var resourceId = ResourceIdBuilder.BuildVirtualMachineResourceId("cloudService", "deploymentName", "roleName");

修改之后，问题解决。

至此，我们已经可以获取到监控指标和监控数据，下一步我们要获取自定义的性能计数器，基于自定义的性能计数器来实现自动伸缩。

周国庆