操作说明:构建指标图来分析相关指标How-to: Build a metrics graph to analyze related metrics

指标顾问中的每个指标都由一个模型单独监视,该模型从历史数据中进行学习,以预测未来趋势。Each metric in Metrics Advisor is monitored separately by a model that learns from historical data to predict future trends. 每个指标都有一个应用于它的单独模型。Each metric has a separate model that is applied to it. 但是,在某些情况下,多个指标可能彼此相关,需要跨多个指标分析异常。In some cases however, several metrics may relate to each other, and anomalies need to be analyzed across multiple metrics. “指标图”可帮助解决此问题。The Metrics Graph helps with this.

例如,如果在不同的指标中具有不同的遥测数据流,则指标顾问将分别对它们进行监视。As an example, if you have different streams of telemetry in separate metrics, Metrics Advisor will monitor them separately. 如果某个指标中的异常导致其他指标中的异常,则在处理事件时,在数据中找到这些关系和根本原因会很有帮助。If anomalies in one metric cause anomalies in others, finding those relations and the root cause in your data can be helpful when addressing incidents. 使用指标图,你可以创建已发现异常的可视拓扑图。The metrics graph enables you to create a visual topology graph of found anomalies.

选择一个指标,将第一个节点放入图中Select a metric to put the first node to the graph

单击导航栏中的“指标图”选项卡。Click the Metrics Graph tab in the navigation bar. 构建指标图的第一步是将节点放入图上。The first step in building a metrics graph is to put a node onto the graph. 选择页面顶部的数据馈送和指标。Select a data feed and metric at the top of the page. 节点将显示在底部面板中。A node will appear in the bottom panel.

选择指标

在现有节点上添加节点/关系Add a node/relation on existing node

接下来,需要添加另一个节点,并指定它与现有节点之间的关系。Next you need to add another node and specify a relation to an existing node(s). 选择现有节点并右键单击该节点。Select an existing node and right click on it. 将出现一个上下文菜单,其中包含多个选项。A context menu will appear with several options.

单击“添加关系”,你将能够选择另一个指标,并指定两个节点之间的关系类型。Click Add relation , and you will be able to choose another metric, and specify the relation type between the two nodes. 你还可以应用特定的维度筛选器。You can also apply specific dimension filters.

添加节点和关系

重复上述步骤后,你将得到一个描述所有相关指标之间关系的指标图。After repeating the above steps, you will have a metrics graph describing the relations between all related metrics. 节点颜色提示Hint on node colors

提示

  • 选择指标和维度筛选器时,图中具有相同指标和维度筛选器的所有节点都将显示为蓝色When you select a metric and dimension filter, all the nodes with the same metric and dimension filter in the graph will be colored as blue .
  • 表示图中指标的未选定节点将显示为绿色Unselected nodes that represent a metric in the graph will be colored green .
  • 如果在当前指标中观察到异常,则该节点将显示为红色If there's an anomaly observed in the current metric, the node will be colored red .

构建指标图时,每当在图中的指标上检测到异常时,你都可以查看相关的异常状态,并获取事件的概要视图。When the metrics graph is built, whenever an anomaly is detected on metrics within the graph, you will able to view related anomaly statuses, and get a high-level view of the incident.

单击图中的事件,并向下滚动到诊断信息下的“跨指标分析”。Click into an incident within the graph and scroll down to cross metrics analysis , below the diagnostic information.

查看相关指标和异常

后续步骤Next steps