Disk full — emergency cleanup and prevention

Summary

If the server ran out of disk space: find large files, clean up logs, temporary files, and set up automatic cleanup.

Applies to:
✔ VPS
✔ Dedicated servers
✔ Linux (Ubuntu, Debian, CentOS)

Diagnostics: how much space is used

Overview by partitions:

df -h

If the root partition / is 95-100% full, you need to free up space immediately.

Finding large files and directories

Find the largest directories:

du -sh /* 2>/dev/null | sort -rh | head -10

Drill into the largest one:

du -sh /var/* 2>/dev/null | sort -rh | head -10

For a convenient interactive view, install ncdu:

apt install ncdu -y

ncdu /

Cleaning up logs

System journal (journald)

Check journal size:

journalctl --disk-usage

Clean entries older than 3 days:

journalctl --vacuum-time=3d

Limit maximum journal size. In /etc/systemd/journald.conf:

SystemMaxUse=200M

systemctl restart systemd-journald

Application logs

Check log sizes in /var/log:

du -sh /var/log/* | sort -rh | head -10

Clear old logs (without deleting the file itself):

truncate -s 0 /var/log/syslog.1

Cleaning temporary files

rm -rf /tmp/*

rm -rf /var/tmp/*

Cleaning package manager cache

Debian/Ubuntu:

apt clean

CentOS/RHEL:

yum clean all

MySQL/MariaDB: binary logs

MySQL binary logs can take up tens of gigabytes:

du -sh /var/lib/mysql/

ls -lh /var/lib/mysql/mysql-bin.*

Remove old binary logs (keep last 3 days):

mysql -e "PURGE BINARY LOGS BEFORE DATE(NOW() - INTERVAL 3 DAY);"

Limit retention in MySQL config (/etc/mysql/my.cnf):

expire_logs_days = 7
max_binlog_size = 100M

Removing old kernels (Ubuntu/Debian)

apt autoremove --purge -y

Prevention: scheduled cleanup

Add a cron job for regular cleanup:

crontab -e

Add a line (clean logs older than 14 days weekly):

0 3 * * 0 find /var/log -name "*.gz" -mtime +14 -delete

Verification

df -h /

Root partition usage should be below 85%.

If space is not freed after cleanup or the disk fills up again within hours, open a support ticket. A disk expansion or investigation into abnormal data growth may be needed.