10.7 - Only one FC path per LUN is registered
Mandy Kirkconnell
I'm not sure if this is the right mailing list for this, but I'm
desperate for answers so here goes....
10.7 only registers 1 FC path per LUN, rather than 1 path per controller
as in 10.6. Does anyone know if 10.7 does its own internal FC
multipathing? If 10.7 is handling multipathing in a transparent manner,
is there any way to disable multipathing? Or at least a way to guarantee
we are using the same physical path each I/O to prevent LUN thrashing?
Ideally, we would like to be able to fine tune the FC paths down to the
WWPN of the FC HBA. For example, if a dual-port FC HBA is connected to a
FC switch, which is connected to both controllers of a RAID unit, we
would want access to all 4 paths:
FC HBA Port 1 -> FC Switch -> RAID Controller A
FC HBA Port 1 -> FC Switch -> RAID Controller B
FC HBA Port 2 -> FC Switch -> RAID Controller A
FC HBA Port 2 -> FC Switch -> RAID Controller B
However, at this point we would even be happy with how things were done
in previous Mac OS X releases, where one path per controller (WWNN) gets
registered:
FC HBA -> FC Switch -> RAID Controller A
FC HBA -> FC Switch -> RAID Controller B
10.7 11A511 (released) only registers one path per LUN, despite the fact
that both controllers are connected:
FC HBA -> FC Switch -> RAID Controller B
This gives us absolutely no way to perform path management of our own.
We are also seeing severe LUN thrashing across the 2 controllers.
This seems like a major regression in 10.7.
If there is a more appropriate mailing list for this topic, please let
me know. I've posted these questions to the Dev Forum, with no response.
I also have a bug report open with Apple, 9625043, also with no response
(~45 days since opening). I really need answers to these questions. At a
minimum, a better understanding of internal multipathing on 10.7, and if
there are any config options to disable it.
Any info, ideas or advise would be very much appreciated!!
Thanks,
Mandy
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Ata-scsi-dev mailing list (Ata-sc...@lists.apple.com)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/ata-scsi-dev/ata-scsi-dev-garchive-72467%40googlegroups.com
This email sent to ata-scsi-dev-...@googlegroups.com
Manoj Radhakrishnan
10.7 does its own FC multipathing. This has changed from 10.6 which did target based multipathing ( based on WWNN ). On 10.7 this has changed to lun based multipathing ( based on the unique lun identifier in the device identification VPD page ).
If the target supports ALUA ( Asymmetric logical unit access ), the host will do IOs only to the target ports configured as "Active-Optimized" for a particular lun. If there is no ALUA support, all paths to a particular lun will be considered active-optimized and IOs will be done across all the paths.
Ideally if the target does not support ALUA, you should configure the luns such that a lun is visible only across the ports of one of the target controllers ( lun affinity ) to prevent lun thrashing.
-manoj
> http://lists.apple.com/mailman/options/ata-scsi-dev/radhakrishnan%40apple.com
>
> This email sent to radhak...@apple.com
Chris Sarcone
------------------
6 Infinite Loop
M/S 306-2MS
Cupertino CA 95014
phone: (408) 974-4033
fax: (408) 862-7577
email: sar...@apple.com
Chris Sarcone
'mpioutil' looks promising for what we need, which is basically the ability to enable/disable specific FC paths.
Does anyone know if the mpioutil source code is available? I've looked in the usual location but can't seem to find it:
www.opensource.apple.com/release/mac-os-x-107/
I'd like to determine if we have access to whatever library or API mpioutil uses, so we can make similar callouts from our own code (rather than calling mpioutil from the command-line). According to 'pkgutil' output, the 'mpioutil' binary is part of com.apple.pkg.BaseSystemBinaries.