Delete old backups

Our database and files backup scripts write a new set of archives every time they run, and nothing ever removes the old ones. In this lesson we'll apply a retention policy and write a small script to delete everything older.

We'll look at how much disk space our backups take up, then use find to remove the ones we no longer need to keep.

Disk space

A site with a 200 MB archive and a 50 MB database dump adds about a quarter of a gigabyte of backups every day. Multiply that by the number of sites on your server and the number of days you'd like to keep, and it becomes clear why a retention policy is needed early on.

Running out of disk space causes all sorts of problems on a WordPress server. Uploads stop working, MariaDB can refuse writes, and logging stops. It's also an annoying problem to fix remotely, so it's much better to avoid it.

Retention

We settled on 14 days of retention in the backup strategy lesson, so anything older than that can go. The find utility can take care of this in a single command:

find /backups -type f -name '*.gz' -mtime +14 -delete

The -type f option limits this to files, -name '*.gz' matches both our .sql.gz dumps and our .tar.gz archives, and -mtime +14 selects the ones that were last modified more than 14 days ago. We never touch our backups after writing them, so their modification time is also the time the backup was taken.

Note that this walks the whole /backups directory, so it covers every site at once, and there's no need to loop through them like we did in the previous two lessons.

The prune script

There isn't much to this one:

KEEP=14

echo "Removing backups older than $KEEP days"
find /backups -type f -name '*.gz' -mtime +$KEEP -print -delete

The -print option lists each file as it's removed, which will give us something useful to look at once these scripts are running from cron.

I called this script prune-backups.sh and placed it in the /config/bin/ directory next to the other two. You can also find it in the config repository on GitHub.

The script needs to run as root to delete files in /backups, so to invoke it as my karl user I'll need sudo:

sudo /config/bin/prune-backups.sh
ls -lh /backups/uncached.org/

I prefer to keep this separate from the backup scripts. Retention lives in one place instead of two, and the script can be tweaked and run on its own whenever disk space is running low.

Testing before deleting

Before running this for the first time, leave out the -delete and have a look at what find picks up:

find /backups -type f -name '*.gz' -mtime +14

Everything listed here is what the script would remove. If the list is longer than you expect, or contains files you don't recognize, it's worth understanding why before running the script itself.

When backups stop

Deleting by age assumes that new backups keep arriving. If our backup scripts stop running for some reason, this script will carry on deleting until there's nothing left.