Yesterday we've attempted a yum update within our Centos 7.4 VM, as we did earlier, using a qcow2 snap, using the main image as backing store, as described in this tutorial:
However, after all was looking good, I shut down the VM, commited the SNAP to the main image, and modified the VM to point to the main store again. After booting up surprise, no longer booting, after hours of debugging our team was able to point out that the XFS filesystem was corrupted and attempted to repair it with xfs_repair.
Managed to repair it, but now we have several problems with broken packages (which we will be able to repair with RPM/YUM).
Is it not safe to commit these snaphots to the source qcow2 image? We've done this in the past and it used to work without problems. I was not aware of any problems that we may have because of this. Did this several times so far.
Can this be realted to the face that the we have an older Centos on the KVM hypervisor than on the Gues operating system? KVM has Centos 7.2, while the VM has Centos 7.4, and both are using XFS filesystem.
Please advise. Did anyone else have this problem in the past?