This is the documentation for Cloudera Manager 4.8.2.
Documentation for other versions is available at Cloudera Documentation.

Charting Time-Series Data

Cloudera Manager enables you to search for a time series, chart the time-series data, group (facet) the individual time series if your search produced multiple time series, and save the results as a view.

The following sections have more details on the terminology used, how to search for time-series data, displaying chart details, editing charts, and modifying chart properties:

Terminology

Entity
A Cloudera Manager component that has metrics associated with it, such as a service, role, or host.
Metric
A property that can be measured to quantify the state of an entity or activity, such as the number of open file descriptors or CPU utilization percentage. For a list of all the metrics supported by Cloudera Manager, see Cloudera Manager Metrics.
Time series
A list of (time, value) pairs that is associated with some (entity, metric) pair. For example, (datanode-1, fd_open), (hostname, cpu_percent). In more complex cases, the time series can represent operations on other time series. For example, (datanode-1 , cpu_user + cpu_system).
Facet
A display grouping of the dataset, shown in separate charts. By default, when a query returns multiple time series, they are displayed in individual charts. Facets allow you to display the time series in separate charts, in a single chart, or grouped by various attributes of the set of time series.

Searching for Time-Series Data

To access the charts search page, in the Cloudera Manager Admin Console, select Charts > Search. You can search for a time series in the Cloudera Manager Admin Console in two ways: by selecting from a list of metrics or by constructing a tsquery statement in the Search text box.

Search Method Procedure
List of Metrics
  1. Click the List of Metrics link to display a list of metrics organized in categories.
  2. Type a name or description of the metric in the Filter by Metric... text box. As you type, only the list of categories containing metrics that match the letters will display.
  3. Click a category to display the metrics related to that category. Note that some metrics will appear in multiple categories as they may apply to more than one entity or activity (for example, the Alerts metric applies to a large number of categories).
  4. Click the metric that you want to chart. The corresponding tsquery statement will appear in the text box and the chart(s) that display that time series will display.
tsquery Statement
  1. Press Spacebar in the Search text box. tsquery statement components display in a drop-down list. These suggestions are part of type ahead, which helps build valid queries. Scroll to the desired component and click Enter. Continue choosing query components by pressing Spacebar and Enter until the tsquery statement is complete.
  2. Click the Search button.

For example, the query SELECT jvm_heap_used_mb where clusterId = 1 could return a set of charts like the following:

Using Context-Sensitive Variables in Charts

When editing charts from a service, role or host status or charts page, or when adding a chart to a status page, a set of context-sensitive variables will be displayed below the query box on the search page. For example, you might see similar variables for the query shown below:

select load_1, load_5, load_15 where entityName=$HOSTID 
$HOSTID = serv-1.my.company.com $HOSTNAME = serv-1.my.company.com  $CLUSTERID = 1

Notice the $HOSTID portion of the query string. $HOSTID is a variable that will be resolved to a specific value based on the page before the query is actually issued. In this case, $HOSTID will become serv-1.my.company.com.

The chart below shows an example of the output of a similar query.

Context-sensitive variables are useful since they allow portable queries to be written. For example the query above may be on the host status page or any role status page to display the appropriate host's swap rate. Variables cannot be used in queries that are part of global views since those views have no service, role or host context.

Displaying Chart Details

When you move your mouse over a chart, its background turns gray, indicating that you can act upon it.

  • Moving the mouse to a data point on the chart shows the details about that data point in a pop-up tooltip.
  • Click a chart to expand it into a full-page view with a legend for the individual charted entities as well more fine-grained axes divisions.
    • If there are multiple elements in the chart, you can check/uncheck the legend item to hide or show that element on the chart.
    • Click the Close button to return to the regular chart view.

Editing a Chart

You can edit a chart from the Custom view and save it back into the same view. Editing a chart only affects the copy of the chart in the current view – if you have copied the chart into other views, those charts are not affected by your edits.

To edit a chart:

  1. Move the cursor over the chart, and click the blue down arrow icon at the top right.
  2. Select Edit. This takes you to the Edit View page for the view you were at, with the chart you selected already displayed.
  3. Click Save to save the revised chart back to the original view.

Copying and Editing a Chart

Editing a copy of a chart is just like editing a chart in the current view, except that you can save it to another existing chart view, back to the current Custom view, or to a new chart view that you create.

To copy and edit a chart:
  1. Move the cursor over the chart, and click the blue down arrow icon at the top right.
  2. Select Edit a Copy. This opens the Chart Search page with the chart you selected already displayed.
  3. Do one of the following:
  4. Click your browser back button to return to your original chart view.

Chart Properties

By default, the time-series data retrieved by the tsquery is displayed on its own chart, using a Line style chart, a default size, and a default minimum and maximum for the Y-axis. You can change the chart type, facet the data, set the chart size, and set X- and Y-axis ranges.

Changing the Chart Type

To change the chart type, click one of the possible chart types on the left: Line, Stack Area, Bar, and Scatter.

Grouping (Faceting) Time Series

A time-series plot for a service, role, or host may actually be a composite of multiple individual time series. For example, the query SELECT jvm_heap_used_mb where clusterId = 1 returns time-series data for the JVM heap used. Each time series has hostname, role type, metric, and entity name attributes. By default each attribute is displayed all on a single chart.

Using facets, you can combine time series based their attributes. To change the organization of the chart data, click one of the facets on the left hand side of the Search screen. The number in parentheses indicates how many charts will be displayed for that facet. As shown in the image below if the serviceName facet is selected for the JVM heap query, the time series is grouped into four charts, one chart each for each service name. The charts for service types with multiple roles contain multiple lines (for example, HBase, HDFS) while services that have only one role (for example, ZooKeeper) contain just a single line.

Changing Dimensions

You can change the size of your charts by moving the Dimension slider. It moves in 50-pixel increments. If you have multiple charts, depending on the dimensions you specify and the size of your browser window, your charts may appear in rows of multiple charts.

The following chart shows the same query as the previous chart, but with All Combined selected (which shows all time series in a single chart) and with the Dimension slider used to expand the chart.

Changing Axes

You can change the Y-axis range using the Y Range minimum and maximum fields.

The X-axis is based on clock time, and by default shows the last hour of data. You can use the Time Range Selector or a duration link () to set the time range. (See Time Line for details).