LSI SAS2308 SCSI controller: Unsupported SA: 0x12

[udotirol]

We're using one of our proxmox nodes as an iSCSI target. The node is equiped with an HP H220/LSI SAS2308 SCSI controller with a number of disks attached to it.

During high load (particularly when cloning or when the backups run), the kernel ring buffer get's spammed with tons of "Unsupported SA: 0x12" messages like below:

[2975074.390610] Unsupported SA: 0x12
[2975074.391782] Unsupported SA: 0x12
[2975074.392881] Unsupported SA: 0x12
[2975074.393836] Unsupported SA: 0x12

I'm talking about more than 1000 messages per second(!) that go on for a longer time, not fitting into the kernel ring buffer.

This looks slightly similar to this thread (in German) https://forum.proxmox.com/threads/probleme-mit-megaraid-sas-8708em2-zfs-over-iscsi.64794/post-328297 but apparently no solution was found there.

Unfortunately google doesn't spit out nothing useful, so maybe someone here has an idea what might be going on?

 

[wolfgang]

Hi,

Proxmox VE does not support the use as an iSCSI target.
What target do you use?
LIO is the better one.

 

[udotirol]

yes, that's what I am using. The proxmox node is serving iSCSI requests using LIO, and to be even more precise, we're talking about ZFS over iSCSI.

 

[wolfgang]

Did you upgrade the Firmware of the Controller to the latest version?
And what firmware do you use on this controller? (HP or LSI, IR or IT)
What kernel do you use?

 

[udotirol]

it runs on the original HP firmware in IT mode, nothing fancy here

But as we use those controllers in a a couple of servers, I noticed that the controller in question has a slightly outdated firmware. I'll update to the latest firmware version coming from HP and report back, if that changes anything.

 

[SBUSSER]

Hi,

I have the same problem with LIO and ZFS over iSCSI, has the firmware upgrade resolved your problem ?

Thank you

 

[aarcane]

This is not related to the SAS controller. This is a bug in the iSCSI target and/or QEMU's iSCSI initiator implementation. I've opened a bug with Proxmox to resolve this. https://bugzilla.proxmox.com/show_bug.cgi?id=4046

 

[strawman]

From this thread https://forum.proxmox.com/threads/zfs-over-iscsi-w-truenas-scale-difficult-troubleshooting.140168/ , I think the messages like kvm: iSCSI GET_LBA_STATUS failed at lba 2695168: SENSE KEY:ILLEGAL_REQUEST(5) ASCQ:INVALID_FIELD_IN_CDB(0x2400) are related.