我们知道,云计算的特点之一是:弹性计算。如果在某一时间点,用于对云计算资源的使用量超过了实际部署的硬件水平,我们就可以通过修改:
但是Windows Azure毕竟是在远程数据中心的托管服务,我们在本地是无法监视和查看Windows Azure托管服务的资源使用情况。如果我们想查看Windows Azure的性能监视器,你可能会想到:
但是这种方法的缺点显而易见:
考虑到这些问题,微软创建了Windows Azure Diagnostics(诊断)的机制。Windows Azure Diagnostics可以诊断检索许多内容,比如跟踪日志,奔溃事件日志等,并且保存下来。在本章里,我将侧重于Peformance Counter(性能计数器),该功能可以提供WIndows Azure应用程序的关键信息。实际上,性能计数器,可以帮助您隔离性能问题。最重要的是,它可以帮您省钱,让Windows Azure的资源得到最佳利用。
Windows Azure Diagnostics
每一个Windows Azure计算节点都有一个内置的诊断进程(DiagnosticsAgent.exe),负责定期收集诊断数据,并缓存到本地文件,并最终存储到一个预定义的Azure存储。请注意,在诊断过程中,也可以手动触发。
具体来说,Peformance Counter的数据被保存到Windows Azure Storage的Table中,表名为WADPerformanceCountersTable。其他Diagnostics,比如跟踪日志,事件日志等数据也都存储在指定的表,像WadLogsTable
,WADDiagnosticInfrastructureLogsTable等
。更多信息,可以参考这里。
每一个Role Instance都有一个Configuration File (配置文件),位于Azure Storage Blob下,Container(概念上类似于文件夹)为wad-control-container。配置文件主要收集性能计数器数据和相关的使用率。
下面显示的是Disnostics Configuration File的位置。你可以下载Cloud Storage Manager访问wad-control-container。
配置文件中使用了一个标准的XML格式,并可以手动修改(不推荐)。
在这篇文章中,我们会研究
1.如果配置Role.cs,并且从外部程序读取Peformance Counter
2.如何访问Azure Storage Table中的数据
3.性能计数器的快速分析
参考自MSDN
为了Diagnostics 你的Role,你必须先在ServiceDefinition.csdef中修改配置文件。具体的您可以参考我的博文Windows
Azure Platform (二十)使用Windows Azure诊断收集日志记录数据
public class WebRole : RoleEntryPoint { /// <summary> /// Entry point before starting the Role /// </summary> public override bool OnStart() { // Get the Role instance Diagnostics configuration. var config = DiagnosticMonitor.GetDefaultInitialConfiguration(); // Define interval for persisting the performance counter data to azure storage. config.PerformanceCounters.ScheduledTransferPeriod = TimeSpan.FromMinutes(5); // CPU Utilization Counter PerformanceCounterConfiguration cpuUtilizationCounter = new PerformanceCounterConfiguration() { CounterSpecifier = @"\Processor(_Total)\% Processor Time", //define the sample internal for the specific performance counter SampleRate = TimeSpan.FromSeconds(1) }; if (!config.PerformanceCounters.DataSources.Contains(cpuUtilizationCounter, new PerformanceCounterComparer())) { config.PerformanceCounters.DataSources.Add(cpuUtilizationCounter); } // Start diagnostic monitor with the new configuration. DiagnosticMonitor.Start ("Microsoft.WindowsAzure.Plugins.Diagnostics.ConnectionString", config); return base.OnStart(); } /// <summary> /// Performance Counter Comparer /// </summary> private class PerformanceCounterComparer : IEqualityComparer<PerformanceCounterConfiguration> { public bool Equals(PerformanceCounterConfiguration a, PerformanceCounterConfiguration b) { //Check whether the compared objects reference the same data. if (Object.ReferenceEquals(a, b)) return true; //Check whether any of the compared objects is null. if (Object.ReferenceEquals(a, null) || Object.ReferenceEquals(b, null)) return false; // Check if the counter specifier and the sampling rate are the same. return (a.CounterSpecifier == b.CounterSpecifier && a.SampleRate == b.SampleRate); } public int GetHashCode(PerformanceCounterConfiguration counter) { //Check whether the object is null if (Object.ReferenceEquals(counter, null)) return 0; //Get hash code for the CounterSpecifier field if it is not null. int hashCounterSpecifier = counter.CounterSpecifier == null ? 0 : counter.CounterSpecifier.GetHashCode(); //Calculate the hash code for the counter. return hashCounterSpecifier ^ counter.SampleRate.GetHashCode(); } } }
使用 Microsoft.WindowsAzure.Diagnostics.Management
.
const string storageAccoutName = "Storage-Name-Here"; const string privateKey = "Storge-Private-Key-Here"; const string deploymentId = "Deployment-Id-Here"; var storageAccount = CloudStorageAccount.Parse(String.Format( "DefaultEndpointsProtocol=https;AccountName={0};AccountKey={1}", storageAccoutName, privateKey)); // Get the diagnostics manager for the entire storage account. var diagnosticManager = storageAccount.CreateDeploymentDiagnosticManager(deploymentId); // Get diagnostics manager for the specific role instance. RoleInstanceDiagnosticManager roleDiagManager = diagnosticManager.GetRoleInstanceDiagnosticManager("WebRole1", "WebRole1_IN_0"); //Modify current configuration var currentConfiguariton = roleDiagManager.GetCurrentConfiguration(); currentConfiguariton.PerformanceCounters.ScheduledTransferPeriod = TimeSpan.FromMinutes(5); currentConfiguariton.PerformanceCounters.DataSources.Add (new PerformanceCounterConfiguration() { CounterSpecifier = @"\Processor(_Total)\% Processor Time", SampleRate = TimeSpan.FromSeconds(1) }); //Commit the changes roleDiagManager.SetCurrentConfiguration(currentConfiguariton);
现在,我们已经配置了我们要监视的计数器。让我们访问Azure Storage Table(表名WADPerformanceCountersTable),并显示出来。
我创建了PerformanceDataContext,派生自TableServiceContext
。这是为了连接到Azure表,Microsoft提供的ADO扩展的一部分。你可以使用LINQ查询以检 索数据。
/// <summary> /// Query helper for retrieving data from the WADPerformanceCountersTable /// </summary> public class QueryExecuter { /// <summary> /// Cloud storage account client /// </summary> private CloudStorageAccount accountStorage; /// <summary> /// Default Constructor - Use development storage emulator. /// </summary> public QueryExecuter() { accountStorage = CloudStorageAccount.DevelopmentStorageAccount; } /// <summary> /// Constructor /// </summary> /// <param name="accountName">Azure storage name</param> /// <param name="privateKey">Azure storage private key</param> public QueryExecuter(string accountName, string privateKey) { accountStorage = CloudStorageAccount.Parse(String.Format( "DefaultEndpointsProtocol=https;AccountName={0};AccountKey={1}", account } /// <summary> /// Retrieve Performance counter data /// </summary> /// <param name="counterFullName">Counter specifier full name</param> /// <param name="deploymentid">Deployment id</param> /// <param name="roleName">Role name</param> /// <param name="roleInstanceName">Role instance name</param> /// <param name="startPeriod">Start sample date time</param> /// <param name="endPeriod">End sample date time</param> /// <returns></returns> public List<PerformanceData> QueryPerformanceCounter(string counterFullName, string deploymentid, string roleName, string roleInstanceName, DateTime startPeriod, DateTime endPeriod) { PerformanceDataContext context = new PerformanceDataContext( accountStorage.TableEndpoint.ToString(), accountStorage.Credentials); var data = context.PerfData; CloudTableQuery<PerformanceData> query = null; query = (from d in data where d.PartitionKey.CompareTo("0" + startPeriod.Ticks) >= 0 && d.PartitionKey.CompareTo ("0" + endPeriod.Ticks) <= 0 && d.CounterName == counterFullName && d.EventTickCount >= startPeriod.Ticks && d.EventTickCount <= endPeriod.Ticks && d.DeploymentId == deploymentid && d.Role == roleName && d.RoleInstance == roleInstanceName select d).AsTableServiceQuery<PerformanceData>(); List<PerformanceData> selectedData = new List<PerformanceData>(); try { selectedData = query.Execute().ToList<PerformanceData>(); } catch { } return selectedData; } }
此演示,我创建了一个WindowsForm,将Diagnostics Table中的数据填充到图表中。
显示的是过去2个小时内,某一个Role Instance的CPU使用率。
Quest 软件公司开发了一个非常方便和容易使用的工具。称为Spotlight on Azure。为我们对整个Azure订阅提供了深度监控的功能,包括Role Instance,数据聚合,历史数据展示,提醒机制,用户自定义深度分析计数器。