Deduplication log displays different values

Published on 15.12.2025


Let’s say the text log shows:

[15.12.2025 01:16:35] - Hardlinks created: 5018 files
[15.12.2025 01:16:35] - Deduplication rate: 99.8% 
[15.12.2025 01:16:35] - Total number of files: 5030
[15.12.2025 01:16:35] - Deduplication rate calculated: 99.8%
[15.12.2025 01:16:35] - Generating detailed deduplication log...
[15.12.2025 01:17:34] - Detailed deduplication log generated.
[15.12.2025 01:17:34] - Deduplication performance: 3937 hardlinks in 59 seconds

At first glance, this looks inconsistent, confusing, and wrong. But it’s correct. Let’s take a closer look.

This is not an error, but rather due to the different ways in which the two numbers are counted.

The explanation

  1. The number 5018 (“Hardlinks created”)

This number is based on the mathematical calculation from the rsync statistics in BmuS:

local hardlink_count=$((GLOBAL_RSYNC_TOTAL_FILES - GLOBAL_RSYNC_TRANSFERRED_FILES))
GLOBAL_RSYNC_TOTAL_FILES 
# Counts everything that rsync has touched. 
# This includes: files, directories (folders), 
# symlinks (shortcuts), device nodes, etc.

If rsync does not need to transfer a folder or file (because it already exists in the reference backup), the script counts this as “saved” or deduplicated.

Important: This also includes 1081 folders and symlinks.

  1. The number 3937 (“Detailed deduplication log”)

This number is generated by the find command, which writes the log file:

find “$backup_dir” -type f -links +1 -printf “%n|%s|%P\n” > “${DEDUP_LOGFILE}.raw”

The parameter -type f is the key. It means: “Search ONLY for regular files.”

Directories are ignored here (technically speaking, directories are not “hardlinked” like files in Linux; they simply exist as a structure).

Symlinks are often ignored here as well (depending on the find implementation and -type f).

The calculation works out.

Statistics (everything): 5018 objects
Log (files only): 3937 files
Difference: 1081 objects

These 1081 missing entries are simply the folder structure and any symlinks (e.g., in /etc/ or /var/www/) that rsync has synchronized (hence in the statistics) but which are not “files with hard links” (hence not in the log).

Conclusion for you:

You don’t need to worry. Your backup is complete. Deduplication works perfectly (99.8%).

The log file intentionally shows only “real files” to remain clear, while the statistics above show the total amount of work saved by the backup.

Previous

Next