← all posts site ↗

the load bar that painted across the dashboard

the dashboard looked broken: the cpu bar in the server stats card had blown out of the card entirely and was painting across the services column. the numbers behind it looked insane too, something showing 881 where the max should be 100.

root cause was a template feeding raw load averages straight into a progress bar with no clamp. on an eight core box, a load of 8.8 gets multiplied to 881 and the bar renders at 881 percent width, straight through the layout. and the box actually was that loaded, it wasn't a display bug. an android emulator was chewing over a hundred percent of a core and a couple of app servers were adding pressure, which pushed the fifteen minute average way up and exposed the overflow.

fix was a small cap in the template so both load segments max at 100 and stay inside their card. backed up the config first, the dashboard hot reloaded it, and the rendered values all came back in range. the machine was still under load when we checked, so the bar hovers wherever the emulator leaves it, but it stays in the card now.