Rack elevation panel

Front and rear views of your racks, unit by unit, drawn from your inventory data and coloured by device state or by any metric.

The placement comes from the data: when the inventory changes (NetBox, Nautobot, a CMDB, a SQL table, a CSV file), the view changes. The panel knows no data source; it reads tables with documented columns, so anything that can produce such a table works without extra code.

• One or several racks side by side, front, rear or both faces, 10-inch or 19-inch, sized to fill the panel.
• NetBox placement rules: lowest unit, descending numbering, full-depth devices, 0U devices.
• Colour, unit, value mappings, data links and variables come from the standard Grafana field config.
• In service, out of service and not monitored states; active alerts badge; free units of each rack; a legend with the counts.
• Device card on hover, data links on click, highlight of a device from a dashboard variable, role filter.
• Overlapping or overflowing devices are drawn and listed, never hidden.

Requirements

Grafana 13.2 or later.

Quick start

Display a rack in a few minutes, with no data source to install:

1. Create a dashboard and add a visualization with the TestData data source.

2. Set Scenario to CSV Content and paste:

   device,rack,position,u_height,face,is_full_depth,role,model,status,value
   patch-hq-01,RACK-HQ-01,42,1,front,false,patch-panel,Panduit CP48,,
   fw-hq-01,RACK-HQ-01,40,1,front,true,firewall,FortiGate 100F,up,41
   sw-hq-core-01,RACK-HQ-01,38,1,front,true,switch,Cisco C9300-48P,up,45
   sw-hq-core-02,RACK-HQ-01,37,1,front,true,switch,Cisco C9300-48P,up,47
   wlc-hq-01,RACK-HQ-01,35,1,front,true,wireless,Aruba 7210,up,62
   wlc-hq-02,RACK-HQ-01,34,1,front,true,wireless,Aruba 7210,down,
   nas-hq-01,RACK-HQ-01,24,2,front,true,storage,Synology RS1221+,up,36
   srv-hq-01,RACK-HQ-01,20,2,front,true,server,Dell R760,up,38
   ups-hq-01,RACK-HQ-01,1,3,rear,true,ups,APC SRT3000,up,29
   pdu-hq-a,RACK-HQ-01,,0,rear,false,pdu,APC AP8853,,
   

3. Pick the Rack elevation visualization.

4. In Standard options, set the unit to Celsius; in Thresholds, add a step at 50 (yellow) and one at 60 (red). The devices take the colour of their temperature.

Data contract

The panel reads table frames: one row per device, one column per attribute. Columns are found by their default names, ignoring case, or mapped to other names in the Columns options.

Devices (required)

Column	Type	Required	Meaning
device	string	yes	Unique device name, the key of the alerts frame and of the highlight variable
rack	string	yes	Rack of the device
position	number	no	Lowest-numbered unit occupied, as in NetBox. Empty for a 0U device
u_height	number	no	Height in units, 1 by default
face	string	no	front or rear, front by default
is_full_depth	boolean	no	Seen from both faces, true by default
role	string	no	Used by the role filter, the label and the card
model	string	no	Shown in the card and the label
status	string or number	no	In service when it is one of the In service values; empty: not monitored
value	number	no	Metric giving the colour, see Colours
parent_device	string	no	The row is listed in the card of its parent instead of being placed

Any other column is shown in the device card, as it is: site, serial number, IP address...

When the frame has a time field, as with a Prometheus range query, only the latest row of each device is read.

Racks (optional)

Column	Type	Meaning
rack	string	Rack, the key of the rack column of the devices
name	string	Name shown above the rack, rack by default
u_height	number	Height in units, 42 by default
starting_unit	number	Number of the first unit, 1 by default
desc_units	boolean	Units numbered from the top, false by default
site	string	Shown under the name
width	number	Width in inches as in NetBox (10, 19...), 19 by default

Without a racks frame, each rack of the devices is drawn with these defaults and grows to fit its highest device. With a racks frame, devices outside the rack are drawn in a dashed overflow zone and listed as problems. The racks are drawn in the order of the Racks option, then of the racks frame, then of the devices.

Alerts (optional)

A frame with a device column and an alertname column, and optionally severity. Each row is an active alert: the device gets a badge with their count, red if one is critical, and the card lists them.

Which frame is which

The frames can be picked in the Data options, by query or frame name. Otherwise the devices frame is the one with device and position or u_height columns, the racks frame the one with rack and no device, the alerts frame the one with alertname.

The panel does not join frames itself. Several queries that return device columns (an inventory and two metrics, for instance) must be joined with the Join by field transformation, mode OUTER (TABULAR), on device. The panel lists the frames it ignores.

Placement

The rules follow NetBox, so that the view matches the rack page of NetBox or Nautobot:

• A device occupies the units from position to position + u_height - 1.
• With desc_units, unit 1 (or starting_unit) is at the top, and a device goes down from its position.
• A full-depth device is drawn on both faces, faded on the opposite face.
• Devices without position, such as PDUs, go to the 0U band right of the rack.
• Two overlapping devices are both drawn; the second one in the data is hatched and takes the right half.
• Problems (overlaps, overflows, duplicates, orphan child devices, unknown racks, ignored frames) are listed under the warning icon in the top right corner. They never stop the drawing.

Colours

State	When	Drawing
Colour of the value	a value, and not out of service	colour of the field config, value in the label
In service, no value	status in service, no value	neutral, green border
Out of service	status not in service	red, label struck through
Not monitored	no status and no value	grey, hatched

The colour of a value is the one Grafana gives it: thresholds, colour scheme and value mappings of the Standard options, with their unit and decimals. A value mapping with a colour on the status field replaces the red of the devices out of service.

Colour fields. An override applies to a column, not to rows, so one column cannot have a threshold in °C for the servers and in amperes for the PDUs. Put each metric in its own column (temperature, current...) and list them in Colour fields: each device takes the first one with a value, and each column gets its own thresholds and unit with an override Fields with name.

Interactions

• Hover a device for its card: inventory, units, state, value, extra columns, active alerts and child devices.
• Click a device to open its data links, configured in the Data links of the field config. The fields of the row are available, for instance https://netbox.example.com/dcim/devices/?q=${__data.fields.device} or /d/my-device-dashboard?var-device=${__data.fields.device}. With one link the click follows it directly.
• Highlight: the devices named in the Highlight variable ($device by default) get a thick border and the others fade. A link from an alert to the dashboard with var-device=<name> lands on the device. A button above the racks clears the highlight.
• Role filter: devices of other roles fade and stop reacting, the racks keep their shape. It accepts a variable such as $role. The legend counts only the devices of the filter.
• Rack links: the name of a rack opens the data links of the racks frame, for instance its NetBox page. The links of the field config defaults apply to every frame and are written for the devices, so give the racks their own with an override Fields returned by query on the racks query, such as https://netbox.example.com/dcim/racks/?q=${__data.fields.rack}.
• Keyboard: Tab moves from device to device and shows the card, Enter opens the links, Escape closes the card and the menu.

Display

• Auto unit height (default): the tallest rack fills the panel height, and the racks scroll sideways when they do not fit in the width. Turn it off to set the unit height in pixels, or use Fit to panel to scale everything down to the panel.
• Under each rack name, the number of free units of the face, a full-depth device of the other face counting as taken.
• The legend at the bottom counts the devices in service (or with a value), out of service and not monitored, and the active alerts.

Recipes

CSV, no infrastructure

Paste the CSV of the quick start in TestData, or serve a CSV file through the Infinity data source. A second query with the racks columns adds rack heights, names and widths.

NetBox with the Infinity data source

Install the Infinity data source and set it up for your NetBox URL, with an Authorization header holding your API token (Token <token>, or Bearer <token> for the v2 tokens of NetBox 4.5+), kept as a secure header.

The device height and depth belong to the device type, which the REST API of devices does not return; the GraphQL API returns them in one query. Add a query of type GraphQL, source URL, URL https://netbox.example.com/graphql/, parser Backend, root selector data.device_list, and the query:

query {
  device_list(filters: { site: { slug: { exact: "$site" } } }) {
    name
    position
    face
    rack {
      name
    }
    role {
      name
    }
    device_type {
      model
      u_height
      is_full_depth
    }
    parent_device {
      name
    }
  }
}

with the columns:

Selector	As	Type
name	device	string
rack.name	rack	string
position	position	number
device_type.u_height	u_height	number
face	face	string
device_type.is_full_depth	is_full_depth	boolean
role.name	role	string
device_type.model	model	string
parent_device.name	parent_device	string

A second GraphQL query, root selector data.rack_list, gives the racks frame:

query {
  rack_list(filters: { site: { slug: { exact: "$site" } } }) {
    name
    u_height
    starting_unit
    desc_units
    width
    site {
      name
    }
  }
}

with name as rack, and u_height, starting_unit, desc_units, width and site.name as site.

Filter and field names can change between NetBox versions: try the queries in the GraphiQL page of NetBox, at /graphql/, first.

State and colour come from your monitoring (Prometheus, Zabbix, LibreNMS...): add a query that returns device and the metric, and join it to the inventory with Join by field on device. The device names of NetBox and of the monitoring usually match on the hostname; otherwise align them with the Rename by regex transformation or, in PromQL, with label_replace.

Prometheus

If an exporter or recording rule publishes the inventory as an info metric, for instance netbox_device_info{device, rack, position, u_height, face, is_full_depth, role, model} 1, three instant queries in Table format give everything:

• A: netbox_device_info
• B: up or any 0/1 health metric with a device label
• C: a metric for the colour, for instance a temperature

Join them with Join by field on device, mode OUTER (TABULAR), then map status to Value #B in the Columns options and set Colour fields to Value #C.

A single query with group_left also works:

up * on (device) group_left (rack, position, u_height, face, is_full_depth, role, model) netbox_device_info

with status mapped to Value. This join keeps only the devices that have the left metric: patch panels and other unmonitored devices disappear. Prefer the three joined queries to see the whole inventory.

Alerts

ALERTS{alertstate="firing", device!=""}

as an instant query in Table format gives the alerts frame. Keep it out of the Join by field transformation with its Apply to filter on the other queries.

Troubleshooting

• No rack to draw: no frame has device and rack columns. Check the column names, or map them in Columns.
• Frame "B" also has a device column and is ignored: join it to the devices frame with Join by field.
• x appears more than once: the same device is on several rows of a frame without time field; only the first one is drawn.
• x overlaps y or x is outside rack: the inventory places two devices on the same units, or past the rack height given by the racks frame.
• The highlight does not follow the variable: write the variable with its $ in Highlight variable.

Feedback and contributing

Report bugs and ask for features in the issues. To work on the code, read CONTRIBUTING.md.

License

Apache 2.0