Why Deleting Apache Logs Doesn’t Free Disk Space and How to Fix It

1. Scenario : A web server running Apache/Tomcat reports full disk space with df, yet after deleting log files the space remains full; du -sh /* shows the total file size far below the disk capacity.

2. Problem analysis : Deleting logs while Apache/Tomcat is active does not free space because the running process still holds the log file’s inode. The file is merely unlinked, so the allocated blocks stay occupied.

3. Solution : Restart Apache/Tomcat, or truncate the log file without deleting it.

4. Extended – better log‑deletion methods :

(1) Delete logs older than three days via a scheduled task, e.g.:

find /data/www/tomcat-webserver/logs/ -type f -mtime +3 | xargs rm -f

find /data/www/tomcat-webserver/logs/*log* -type f -mtime +3 | xargs rm -f

find /data/www/tomcat-webserver/logs/ -type f -mtime +3 -exec rm -f {} \;

(2) Clear the current log without removing the file:

cat /dev/null > ./logs/test.log

> ./logs/test.log

5. Production simulation :

Install Apache and modify the log path:

yum install httpd

sed -i 's#CustomLoglogs/access_log combined#CustomLog /app/logs/access_log combined#g' /etc/httpd/conf/httpd.conf

Create a large file to simulate a small‑capacity disk:

dd if=/dev/zero of=/mnt/apachelog bs=10K count=10

mkfs.ext4 /mnt/apachelog

Mount the loop device to the log directory:

mkdir -p /app/logs

mount -o loop /mnt/apachelog /app/logs/

Start Apache and generate heavy traffic to fill the disk:

/etc/init.d/httpd start

for i in $(seq 10000); do curl -s http://192.168.1.200; done

Simulate normal log deletion:

rm -f /app/logs/access_log

df -h

The df -h output shows the filesystem still reports high usage, and lsof | grep access_log reveals processes (httpd) still holding the deleted log file.

6. Final resolution :

(1) Restart Apache to release the inode: /etc/init.d/httpd restart (2) Truncate the log file directly:

> /app/logs/access_log