On 05/02/2024 02:02, Jan Palus wrote:
I've tried to replicate the env (i686 VM, XFS) but without success. Not
sure if it's block device driver? XFS? or something else? Can't do much
more without ability to reproduce. Any ideas are welcome.
All builders run on the same physical machine as VMs under proxmox 8.1:
pve-manager/8.1.4/ec5affc9e41f1d79 (running kernel: 6.5.11-8-pve)
Each builder uses two physical SSD disks (crucial: x32 - mx200, i686 -
mx300, x86_64 - mx500).
SSD are available to VMs as passthru. VMs have soft RAID1 setup inside,
xfs fs on top of that. It's basically old hardware setup migrated from 3
different physical machines into one with proxmox.
Providing disk directly to VM seems to be a key to reproduce the problem
(as I was unable to reproduce it on the same machine when VM visible
disk was backed by just qcow2 file).
Disks used default options (so no cache, no discard, no io thread, no
skip replication, async io - io_uring). Virtio interface.
But I noticed that enabling discard on these hard disks (in
proxmox->VM->Hardware->Hard disk) makes problem go away.
So enabled on all builders.
--
Arkadiusz Miśkiewicz, arekm / ( maven.pl | pld-linux.org )
_______________________________________________
pld-devel-en mailing list
pld-devel-en@lists.pld-linux.org
http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
