Arm
3 boards ran 133,383 evaluations in the last 12 months. 3,862 first attempts failed; 741 passed on retry. 3,045 evaluations still failed after retries, most on Musca-B.
The bench
One tape per board: evaluations per week, stacked by final outcome. Each tape has its own scale, topped by its busiest week, so compare boards by the totals beside their names. Failures sit on top as caps.
Changes with a failed boot on this board
Distinct change and patchset pairs whose evaluation on the board (Allow-CI+2 for TF-A, linked runs for TF-M) still failed its boot check after retries, out of pairs with a boot verdict. Only runs whose Jenkins lineage was captured can count, so each board shows the share of its evaluations linked to a change. Gerrit votes are not collected, so this is not a count of blocked changes.
Failed after retries
CI evaluations with a final code, lab or unclassified failure, newest first; 25 of 3,045 shown. Health checks and manual runs are never listed here.
Fleet health
Health checks from the last 7 days, whatever the window above. They are not CI results and never count towards any rate on this page. Check units means at least one unit's latest health check failed.
15 devices across 3 board types. Devices are units that ran any LAVA job, health checks included, in the last 30 days, whatever the window or project above. Every board type that is not retired counts, including those with health checks only.