I/O errors with NVM drives

[luison]

Current kernel used: 5.3.13-3-pve

I seem to having some ocassionall I/O errors issues with my NVM pci-e 4 drives (Corsair MP600 on a ASUS Pro WS X570-ACE motherboard) that I've come to determine that could be related to some kernel trim support of NVM drives as reflected on kernel bug:

https://bugzilla.kernel.org/show_bug.cgi?id=202665

Errors like:
Feb 7 22:01:13 e20home kernel: [32434.453698] blk_update_request: I/O error, dev nvme1n1, sector 14203336 op 0x3DISCARD) flags 0x800 phys_seg 1 prio class 0

It shows up sometimes when deleting or creating partitions and can be forced by executing:
fstrim -v /partition_name

I understand by the bug that issue might be resolved on kernel version 5.4.7 or via patching the current kernel.
Wondering if anyone having similar issues, any possible fixes other than patching the kernel and/or how soon to expect the PVE to reach that kernel version.

Thanks.

 

[t.lamprecht]

I understand by the bug that issue might be resolved on kernel version 5.4.7 or via patching the current kernel.
Wondering if anyone having similar issues, any possible fixes other than patching the kernel and/or how soon to expect the PVE to reach that kernel version.

FYI, the patch mentioned in the bug report you linked is included in the pve-kernel-5.3.18-1-pve kernel which is available through our no-subscription repository since today. Please update and see if that helps.

 

[luison]

That's great news, thanks.
I managed to update the NVMs MP600 firmware (via Windows PE rescue) and the problem seem to have stopped (reporting at least).
Update to latest kernel now and all seems working fine. Thanks.

 

[t.lamprecht]

Great to hear, thanks for your feedback!