Filesystem usage is the space consumed and remaining on each mounted filesystem, reported per mount point along with the backing device and the filesystem type. Of all the ways a server fails, this is the most predictable and the easiest to catch early.
A static percentage threshold makes a weak alert. 90% full on a 20 GB volume leaves 2 GB, which one chatty log file eats in an afternoon, while 90% on a 4 TB volume leaves 400 GB and needs no attention at all. Alert on absolute free bytes together with the rate of fill, so the page arrives with hours to act rather than at the moment writes start failing. Two more details are worth knowing: pseudo filesystems such as tmpfs, proc, and overlay layers clutter the list and should be filtered out, and a filesystem can exhaust its inode table while free space still looks fine, which is why "no space left on device" sometimes contradicts the graph.
Full disks fail in ways that look like unrelated bugs: a database refusing writes, logs silently truncated, a package manager that will not run. Hyperping reports used and free space per real mount point with the device, mount point, and filesystem type attached, and filters pseudo filesystems out before the data reaches the dashboard.
