Debugging boots and reboots?

• When the rootfs is unclean, fsck often requires a reboot, so to avoid a double reboot, we pivot_root to a tmpfs, unmount the root partition, fsck it, mount it back, and pivot_root back to it.

• We have serial console on some test machines, but not worth the price on all machines.

• bootlogd will capture boot messages without requiring the much heavier and buggy plymouth.

• Each initscript runtime is logged, so we can find which ones can take too long.

• We also know if we booted with missing RAM, or a dead disk, and so forth.