[DRBD-user] Another iSCSI Active/Active Thread

Myles Gray myles.gray at exitex.com
Thu Oct 10 09:36:04 CEST 2013

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


That's very clear thanks for your input Lars!

Regards,
Myles
________________________________________
From: drbd-user-bounces at lists.linbit.com [drbd-user-bounces at lists.linbit.com] on behalf of Lars Ellenberg [lars.ellenberg at linbit.com]
Sent: 09 October 2013 19:39
To: drbd-user at lists.linbit.com
Subject: Re: [DRBD-user] Another iSCSI Active/Active Thread

On Wed, Oct 09, 2013 at 03:33:51PM +0000, Myles Gray wrote:
> I know that this has been discussed many times over by the community
> and Linbit engineers, however, is iSCSI active/active now safely
> deployable on DRBD using the built in kernel scsi target (LIO)?
>
> I ask because it now supports VMWare's VAAI which is nothing more than
> a collection of t10 SCSI commands (it's not an API at all, why they
> called it that I don't know):
>
> 0x93    WRITE SAME(16)
> 0x41    WRITE SAME(10)
> 0x42    UNMAP
> 0x89    SCSI COMPARE and WRITE - ATS
> 0x83    EXTENDED COPY
>
> The SCSI COMPARE AND WRITE command is used by clustering aware
> applications for resource locking, for example ESXi - therefore if
> your initiator is ESX's iSCSI Initiator and your target (LIO) supports
> VAAI then you can run active/active iSCSI on DRBD with no problems?

The basic problem is that targets keep state (iSCSI is not stateless).
But that state is neither communicated nor coordinated between multiple
such targets in any fashion (or I missed some recent development).
Which is why things *will* break if you want to talk to two independend
such targets concurrently. (Even though you can likely build setups that
will *appear* to work for a long time -- in fine weather).

Whether or not those target implementations learn to support
additional features is unrelated to this particular problem.

The capabilities and features of *one* single target implementation
have absolutley nothing to do with the fact that
you want to talk to *two* independend such targets, concurrently.
And they don't know and don't care about each other.

It does not help that the target now supports commands that can be used
by cluster aware applications.
If you want to use multiple targets concurrently on the same data set
in a cluster, the target needs to become cluster aware *itself*.

Whether or not the backend storage is replicated by DRBD,
shared SCSI or SAN, or whatever else, is unrelated as well.

--
: Lars Ellenberg
: LINBIT | Your Way to High Availability
: DRBD/HA support and consulting http://www.linbit.com

DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.
__
please don't Cc me, but send to list   --   I'm subscribed
_______________________________________________
drbd-user mailing list
drbd-user at lists.linbit.com
http://lists.linbit.com/mailman/listinfo/drbd-user



More information about the drbd-user mailing list