Multi Dimensional Dashboard Variables

Introduction: The Need for Multi-Dimensional Filtering

In the previous lesson, you created your first query variable — a host selector that filters all panels on your dashboard. This was a significant improvement over hard-coded queries because it allowed you to switch between different hosts (host-a, host-b, host-c) with a single dropdown selection. However, if you look closely at your metrics data, you'll notice that host filtering alone isn't always enough to create focused visualizations. Consider your disk metrics table. Each host doesn't just have one disk — it has multiple mount points. host-a might have a root partition (/), a var directory (/var), and a data volume (/data). When you filter by $host, your disk usage panel shows all three mount points combined, making it difficult to focus on a specific partition. The same issue exists with network metrics: each host has multiple network interfaces like eth0 and eth1 . Filtering by host alone shows traffic across all interfaces mixed together. What you need is the ability to filter along multiple dimensions simultaneously. You want to say, "Show me disk usage for host-a's root partition," or "Show me network throughput for host-b's eth0 interface." This requires combining your existing host variable with new variables that capture these additional dimensions. In this lesson, you'll create two new variables — one for disk mount points and one for network interfaces — and learn how to use multiple variables together in your WHERE clauses.

Creating the Mount Variable

Start by creating a variable that lists all available mount points from your disk metrics. This follows the exact same pattern you used when creating the host variable in the previous lesson: you’ll write a query that returns distinct mount points, and Grafana will use those results to populate a dropdown menu. Open your dashboard settings by clicking the gear icon in the top right, select Variables from the sidebar, and click Add variable . Configure the variable with the following settings: Name: mount, Label: Mount Point, Type: Query, Data source: Postgres Local. Then enter this query: SQL SELECT DISTINCT mount AS text, mount AS value FROM public.metrics_disk ORDER BY mount; This query retrieves all unique mount points from your disk metrics table. The DISTINCT keyword ensures each mount point appears only once, even though it appears in hundreds or thousands of rows across different timestamps. The AS text, mount AS value pattern should be familiar from the previous lesson — this is the format Grafana requires for query variables. The text column determines what users see in the dropdown (like /, /var, /data), and the value column determines what gets substituted into your queries when that option is selected. The ORDER BY mount clause sorts the mount points alphabetically. This is particularly useful for mount points because the forward slash character sorts before letters, which means your root partition (/) will always appear first in the list, followed by other paths like /data and /var .

Join the 1M+ learners on CodeSignal

Be a part of our community of 1M+ users who develop and demonstrate their skills on CodeSignal

Testing Multi-Dimensional Filtering

Now that you understand the query structure, let's see how these variables work together in practice. When you view your dashboard, you'll see three dropdowns at the top: Host, Mount Point, and Interface. These selectors are independent but work together when your panel queries reference multiple variables.

Suppose you create a disk usage panel using the query from the previous section. When you first load the dashboard, it might default to showing host-a, /, which displays disk usage for host-a's root partition. The panel shows a single time series line representing the used_pct values over time for that specific combination.

Now click the Mount Point dropdown and select /data instead. The moment you make this selection, the panel automatically re-queries the database. The query still filters by host-a (because that's still selected in the host dropdown), but now it's looking for rows where mount = '/data'. The visualization updates to show the disk usage of the data volume instead of the root partition. You've just switched from monitoring one partition to another without editing any queries or creating duplicate panels.

This is the power of multi-dimensional filtering. Your disk metrics table might contain thousands of rows covering three hosts, each with three mount points, recorded every minute over weeks of time. That's potentially over 100,000 rows of data. But with two dropdown selections — picking a host and a mount point — you can instantly zoom in on exactly the subset of data you need to see. Instead of looking at nine time series (three hosts times three mounts), you're looking at just one: the specific mount point on the specific host that matters to you right now.

The same interaction works for network panels. If you have a network throughput panel using the interface variable, selecting eth0 shows traffic for that interface, while selecting lo shows loopback traffic. Each dropdown selection applies an additional filter, narrowing the results further. You could have dozens of hosts with multiple interfaces each, but with just two clicks, you can focus on a single interface on a single host.

What makes this approach scalable is that you don't need to create separate panels for each combination. Instead of building 27 panels (3 hosts × 3 mounts × 3 interfaces), you build 2 panels and add 2 variables. As your infrastructure grows — adding more hosts, more mount points, or more network interfaces — your dashboard continues to work without modification. The variable queries automatically pick up new values from your data, and your panels dynamically filter to whatever combination you select.

Introduction: The Need for Multi-Dimensional Filtering

Creating the Mount Variable

Join the 1M+ learners on CodeSignal

Creating the Interface Variable

Combining Variables with AND in WHERE Clauses

Combining Variables with AND in WHERE Clauses

Testing Multi-Dimensional Filtering

Summary and Practice Preview