fsarchiver hangs on Fedora 35/36 on Hyper-V

[mtniehaus]

It's a weird issue: We are using fsarchiver on Fedora 35/36 to restore a Redhat root image. Close to 80% of the time, the restfs process will hang. It appears to be waiting on IO (and we see a corresponding wait in xfsaild, so it looks like those processes are effectively deadlocked). The exact point of the hang varies (might be 5% into the FS, might be 75%, or anywhere in between).

But this hang is only happening on Hyper-V, it's working fine on physical machines and in Proxmox.

The problem also only happens when applying an xfs file system. If we override that to specify to create an ext4 file system, it works fine (but then RHEL is unhappy later), with no hang during the restfs.

Any thoughts on how to get around this?

[marcosfrm]

Since it works with EXT4 sounds like a kernel bug. You can try strace fsarchiver.

RHEL still supports EXT4, you just need to add the module to the initramfs. Take a look here:

https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/9/html-single/deploying_red_hat_enterprise_linux_9_on_public_cloud_platforms/index#installing-hyperv-drivers_configure-the-image-azure

https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/9/html/deploying_red_hat_enterprise_linux_9_on_public_cloud_platforms/assembly_deploying-a-virtual-machine-on-aws_cloud-content#completing-the-rhel-installation_deploying-a-virtual-machine-on-aws

(add_drivers+=" xxx " and dracut -f --add-drivers xxx parts)