Reports

Introduction

The collector embeds a reporting framework.

Reports are most useful to extract key point indicators and billing metrics from the collector database.

Overview

A report is a collection of charts and metrics.

Metrics are datasets extracted, daily or just-in-time, from the collector database using a user-defined SQL request.

Reports are presented to users through the Statistics ‣ Reports menu, and can be accessed directly using each report's direct link.

Privileges

Report, charts and metrics creation requires the ReportManager privilege.

Testing

Report, charts and metrics definitions can be tested individually using the gear-icon on the definition line of their admin view.

Metrics

The dedicated view is Admin ‣ Metrics.

Properties

A metric is defined by the following properties:

  • Name

    A name to describe the metric.

  • SQL request

    The SQL request used daily to extract and store the dataset.

  • Value column index

    If the SQL request has more than one column, set this property to the index of the column containing the value to historize.

  • Instance column index

    If the SQL request extract a grouped-by dataset, the user will want to include a column containing each group description. This property must then be set to the index of this column. Such dataset are extracted at report rendering, and never historized. We will refer to this type of metric as tabbed metrics.

Honoring filtersets

Filtersets with the stats property set can influence a metric dataset. For example, a metric dedicated to nodes counting would extract the node count for each stats-filterset, and having one of those filtersets as a session filter will make to metric show the dataset for this active filterset.

This can be achieved using in the SQL request the special keywords %%fset_nodenames%% or %%fset_services%%.

Be aware of the collector resource usage caused by this feature. Many stats-filtersets coupled with under-optimized metric SQL-requests can load the collector for a long period each day.

Of note, a filterset has the stats property off by default.

Examples

  • Historized metric

    Name: compliance fixes per day SQL request: select count(id) from comp_log where run_action="fix" and run_date > date_sub(now(), interval 1 day) and run_nodename in (%%fset_nodenames%%) Value column index: 0 Instance column index: (empty)

  • Tabbed metric

    Name: nodes per city SQL request: select count(id), loc_city from nodes where nodename in (%%fset_nodenames%%) group by loc_city Value column index: 0 Instance column index: 1

Charts

The dedicated view is Admin ‣ Charts.

Properties

A chart is defined by the following properties:

  • Name

    A name to describe the chart.

  • YAML definition

    The chart definition in YAML format, where the user can point the metrics to aggregate in the chart and set properties like metrics stacking. The structure of the expected YAML is detailled in the next section.

YAML definition

The expected YAML structure is a dictionnary with the following keys.

Title

A string describing the chart. This string will be rendered with the chart, unlike the Name chart property explained in the previous section.

Options

A dictionnary structure supporting the following keys.

  • stack

    Candidates: Yes, No

    Wether the embedded metrics should be stacked one over the other, or rendered unstacked.

Metrics

A list structure. Each member is a dictionnary structure pointing to a metric defined in the Admin ‣ Metrics view, using the following keys.

  • metric_id

    Type: integer

    The metric Id as reported in the Admin ‣ Metrics view.

  • label

    Type: string

    The label representing the metric in the chart legend.

  • fill

    Candidates: Yes, No

    Fill the zone under the metric line. To set only for one metric if stacking is disabled.

  • shadow

    Candidates: Yes, No

    Render a shadow under the metric line or filled-in zone.

  • unit

    Type: string

    The unit to use when formatting the metric instance values. Example: "MB". Defaults to an empty string.

Example YAML definition

Title: Nodes per city

Options:
 stack: no

Metrics:
 -
  metric_id: 39
  fill: no
  shadow: yes

Reports

The dedicated view is Admin ‣ Reports.

Properties

A report is defined by the following properties:

  • Name

    A name to describe the report.

  • YAML definition

    The report definition in YAML format, where the user can point the metrics and charts to aggregate in the report and set rendering properties. The structure of the expected YAML is detailled in the next section.

YAML definition

The expected YAML structure is a dictionnary with the following keys.

Title

A string describing the report. This string will be rendered with the chart, unlike the Name chart property explained in the previous section.

Sections

A dictionnary structure, supporting the following keys Title, Desc and children.

Keys details:

  • children

    A list of dictionnary structures. Each dictionnary can describe a chart or a metric

    • Chart keys: Title, Desc and chart_id

    • Metric keys: Title, Desc and metric_id

    • width

      The width in percent of the section element. Of note the padding and margin of elements are set to 1%, so the width must compensate for this, substracting 4% to each element width. For example to render a chart and a metric in 2/3 - 1/3 layout, the width should be set to 62% and 29%, the remaining 8% being used by 4 * 1% margins and 4 * 1% paddings.

  • Title

    A string describing the metric or chart.

  • Desc

    A string detailling or commenting the metric or chart. Optional.

  • chart_id

    Type: integer

    The chart Id as reported in the Admin ‣ Charts view.

  • metric_id

    Type: integer

    The metric Id as reported in the Admin ‣ Metrics view.

Example YAML definition

Title: My report

Sections:
  -
    children:
      -
        Title: Linux release dispatch
        metric_id: 48
        width: 100%

  -
    Title: RAM
    Desc: All about memory
    children:
      -
        Title: Average RAM per node
        Desc: in megabytes
        metric_id: 49
        width: 62%
      -
        Title: Total RAM
        chart_id: 3
        width: 29%