linux - df command says disk 100% used, but it actually it's not - linux

Question

I have an EC2 instance running jenkins service and almost everyday it stops working saying the disk is full.
So I enter the instance to check and it really says 100% used
$ df -h
Filesystem Size Used Avail Use% Mounted on
devtmpfs 987M 60K 987M 1% /dev
tmpfs 997M 0 997M 0% /dev/shm
/dev/xvda1 32G 32G 0 100% /
So I use NCDU to check what is occuping the space and it says only 8gb is used.
ncdu 1.10 ~ Use the arrow keys to navigate, press ? for help
--- / -------------------------------------------------------
3.7GiB [##########] /var
1.7GiB [#### ] /home
1.6GiB [#### ] /usr
1.0GiB [## ] swapfile
323.6MiB [ ] /opt
133.8MiB [ ] /lib
47.3MiB [ ] /boot
43.8MiB [ ] /root
23.1MiB [ ] /public
19.8MiB [ ] /lib64
12.3MiB [ ] /sbin
10.7MiB [ ] /etc
7.0MiB [ ] /bin
3.7MiB [ ] /tmp
60.0KiB [ ] /dev
e 16.0KiB [ ] /lost+found
16.0KiB [ ] /.gnupg
12.0KiB [ ] /run
e 4.0KiB [ ] /srv
e 4.0KiB [ ] /selinux
e 4.0KiB [ ] /mnt
e 4.0KiB [ ] /media
e 4.0KiB [ ] /local
e 4.0KiB [ ] /cgroup
. 0.0 B [ ] /proc
0.0 B [ ] /sys
0.0 B [ ] .autorelabel
0.0 B [ ] .autofsck
Total disk usage: 8.6GiB Apparent size: 8.6GiB Items: 379695
Then I reboot the instance and the free space comes back to only 28% used
$ df -h
Filesystem Size Used Avail Use% Mounted on
devtmpfs 987M 60K 987M 1% /dev
tmpfs 997M 0 997M 0% /dev/shm
/dev/xvda1 32G 8.7G 23G 28% /
What can be causing this problem?
Additional data:
PSTREE
init─┬─acpid
├─agetty
├─amazon-ssm-agen───6*[{amazon-ssm-agen}]
├─atd
├─auditd───{auditd}
├─crond
├─dbus-daemon
├─2*[dhclient]
├─java─┬─sh───sudo───gulp --prod───9*[{gulp --prod}]
│ └─67*[{java}]
├─lvmetad
├─lvmpolld
├─6*[mingetty]
├─rngd
├─rpc.statd
├─rpcbind
├─rsyslogd───3*[{rsyslogd}]
├─2*[sendmail]
├─sshd───sshd───sshd───bash───pstree
├─supervisord───2*[php]
└─udevd───2*[udevd]
CRONTAB
* * * * * cd /home/ec2-user/laravel-prod && php artisan schedule:run >> /dev/null 2>&1
0 1 * * * rm /var/log/jenkins/jenkins.log