SCSI Behavior for I/O Rejected Because of QoS

Gerry Houlder gerry.houlder at seagate.com
Fri Jun 27 08:36:46 PDT 2014


Attachment #1: <a href="http://www.t10.org/cgi-bin/ac.pl?t=r&f=r1406270_nameless-3564-2-1.html">nameless-3564-2-1.html</a>

David's suggestion is good.
There is also work going on for a method to apply a time limit to commands,
and reject commands that cannot be completed within the indicated time
limit. This technique will use 2Eh 00h, 2Eh 01h, 2Eh 02h, or 2Eh 03h as the
additional sense (these are various "insufficient time" or "command
timeout" descriptions) along with ABORTED COMMAND sense key. I think any of
these combinations could be a correct description of your situation.
On Thu, Jun 26, 2014 at 11:30 AM, Lee, Shuyu <shuyu.lee at emc.com> wrote:
> * From the T10 Reflector (t10 at t10.org), posted by:
> * "Lee, Shuyu" <shuyu.lee at emc.com>
> *
> Dear T10-Stds Colleagues,
>
> The VNX block team is working on Multi-Tenancy QoS support for VNX2. In
> this context, IOs may be rejected because of QoS limits, but rejected I/Os
> should be retried.  How should initiators be informed of these IO
> rejections?
>
> David Black suggested using CHECK CONDITION (02h) with ABORTED COMMAND,
> SYSTEM_RESOURCE_FAILURE, INSUFFICIENT RESOURCES (0Bh/55h/03h) and cautioned
> that using a different SCSI status like BUSY or TASK SET FULL may have
> undesirable side effects on some initiators.	He also suggested asking this
> list for additional suggestions and input.
>
> What would you suggest?
>
> Best regards,
> -----------------------------------------------------------
> Shuyu Lee
> EMC Corporation,	  228 South St., Hopkinton, MA	01748
> +1 (508) 249-7610			    shuyu.lee at emc.com<mailto:
> shuyu.lee at emc.com>
> -----------------------------------------------------------
>
>
> *
> * For T10 Reflector information, send a message with
> * 'info t10' (no quotes) in the message body to majordomo at t10.org
>



More information about the T10 mailing list