After upgrading the kernel package, my Centos 7 server didn't boot.
Here is a small summary: old kernel = kernel-3.10.0-123.4.4.el7.x86_64 (everything just fine) new kernel = kernel-3.10.0-123.6.3.el7.x86_64 root partition is on a primary partition, NOT on a LVM volume /var partition is on a LVM volume new kernel's boot problem = systemd cannot mount /var partition saying requirements failed or something like that
I have find out that the new kernel's initramfs image does not contain the lvm module while the old kernel's initramfs did.
I have created a new initramfs with lvm using dracut and could boot the server successfully. However I do not consider this as a fix, and don't know how to find the primary cause of the problem. I need some help finding out who's to blame:
Update: I did some tests on a very similar Centos 7 server and could reproduce the problem with an older kernel. So kernel version doesn't matter.
After swapoff -a the mkinitrd command produces an initamfs image WITHOUT the lvm. The system cannot boot with it.
Then after swapon -a the mkinitrd creates an image WITH the lvm module.
Any ideas why? Is it a bug?
On the original system was the swap not disabled, but replaced with an encrypted swap (one added line in crypttab, one line changed in fstab). And when the next kernel update came, the server have become unusable as I described in my first post.