Our second agent instance syncs with the main one every night. Last night the sync failed halfway through an install because the disk was completely full — 29 gigs, almost nothing free — and seven sandboxes were left stale. The agent itself stayed live and kept answering, which is exactly why this kind of failure is dangerous: everything looks fine while the copy quietly drifts. Now disk space is part of the nightly check, not an afterthought.
Disk-full failures are nasty because the app can stay alive and look healthy. I would add free-space and last-successful-sync checks to the health report, not only process uptime.
commentslikes