After a Debian 9.9 upgrade I restarted the server to load the new kernel image, Debian 4.9.168-1+deb9u5 (2019-08-11) x86_64 GNU/Linux, and it didn't boot anymore.
The server has 4 ssd disks, the main one loads the operating system and mounts correctly:
device fs_type label mount point
---------------------------------------------------------------------------------
/dev/sda1 (not mounted)
/dev/sda2 ext4 /boot
/dev/sda3 swap [SWAP]
/dev/sda4 ext4 /tmp
/dev/sda5 ext4 /
But the other 3 disks, which are mounted in web pages directory (/var/www/virtual/website), generate errors of the type:
Job var-www-virtual-domain1.mount/start failed with result 'dependency'.
This causes the system not to load correctly, the ssh does not load and I have to enter through the console and manually start the service (service ssh start), in order to be able to access using ssh.
Another thing that happens is that once I enter the system, if I try to manually mount the disks in their path, it does not work (nor generates errors), for example, this doesn't work y I have the same mount point in fstab:
mount -t ext4 /dev/sdb1 /var/www/virtual/domain1
But if I create a new directory, it mounts without problems:
mkdir /var/www/virtual/test
mount -t ext4 /dev/sdb1 /var/www/virtual/test
I have managed to fix it by deleting the disks from the fstab and mounting them from a script that I start in crontab with "@reboot root /root/mount.sh", that is, if I start without the disks in fstab and mount them manually, everything works correctly.
This is what I deleted from the fstab to fix it:
# /dev/sdb1
UUID=c5d69d49-c169-4e50-ab51-90bd50c0bc0e /var/www/virtual/domain1 ext4 errors=remount-ro 0 0
# /dev/sdc1
UUID=74af1317-078a-425b-b9e3-2d2f7159d1d9 /var/www/virtual/domain2 ext4 errors=remount-ro 0 0
# /dev/sdd1
#UUID=cd810a2f-080d-4e26-917b-bd6dac59adf4 /backup ext4 errors=remount-ro 0 0
The system had always worked loading the disks from fstab until I did the last operating system upgrade (apt upgrade).
I have 8 other servers with almost identical configurations (different hardware), which have not had this problem with the same update.