In the recent versions of Linux, including Opensuse Leap 15.6, the KVM system is causing host disk filesystem corruption. It happens whenever the disk caching mode is set to Directsync. I have trapped this condition including when a KVM vm is set up to use a SATA virtual drive. The host is using BTRFS.
Let me emphasize the seriousness of this fault. It causes main host filesystem corruption that renders the server’s disk destroyed. This has caused enormous problems in the field, because it wrecks servers which then need to be entirely rebuilt from ground up to be put back in service.
The fault is repeatable and happens every time Directsync mode is selected. Hypervisor Default, or writeback, writethrough do not display it. Here is one report:
~> sudo btrfs device stats /
[/dev/nvme0n1p1].write_io_errs 0
[/dev/nvme0n1p1].read_io_errs 0
[/dev/nvme0n1p1].flush_io_errs 0
[/dev/nvme0n1p1].corruption_errs 7
[/dev/nvme0n1p1].generation_errs 0
This Directsync caching option has a kernel fault associated with it, and in the interim, it must be taken off the Virtual Machine Manger menus at the very least. We need action taking over this urgently. Can anyone please recommend a course of action and alerts over this fault?