We've just installed a new FC SAN and I've just got a couple of questions
about some issues we're seeing.

Basically we've got single ported dual controller's connected to 2 switches.
Each host has a single HBA connected to a switch.  The switch is fabric
enabled and both are interconnected.

Our issue is that our Windows servers (2k & 2k3), can see duplicates of
their own storage.  i.e.: we have created a LUN for host1 that's 300 GB in
size.  Host 1 can see 2 instances of this storage and is able to write to
both instances of storage.

We think this is due to both controllers advertising the presence of the LUN
and our host's HBA being permitted to access it, is seeing it as 2 different
LUN's.  Is this likely to be the cause, or is it possibly something to do
with the cross-connect in the switches?

Is there a way to perhaps force the HBA to only see 1 instance of the
storage, but still allow the LUN to be available when a controller fails
over?

Should mention that the san is an NEC, the switches Brocade Silkworm's and
the HBA's are Emulex's.

Any help would be appreciated.

Dwayne.

[ Post a follow-up to this message ]

dwayne wrote:
> We've just installed a new FC SAN and I've just got a couple of
> questions about some issues we're seeing.
>
> Basically we've got single ported dual controller's connected to 2
> switches. Each host has a single HBA connected to a switch.  The
> switch is fabric enabled and both are interconnected.
>
> Our issue is that our Windows servers (2k & 2k3), can see duplicates
> of their own storage.  i.e.: we have created a LUN for host1 that's
> 300 GB in size.  Host 1 can see 2 instances of this storage and is
> able to write to both instances of storage.
>

All things being equal, with nothing done to prevent it, this would be
normal behaviour, i.e. both controllers present the LUN and when the HBA
initializes it sees two instances of the LUN.

> We think this is due to both controllers advertising the presence of
> the LUN and our host's HBA being permitted to access it, is seeing it
> as 2 different LUN's.  Is this likely to be the cause, or is it
> possibly something to do with the cross-connect in the switches?
>
> Is there a way to perhaps force the HBA to only see 1 instance of the
> storage, but still allow the LUN to be available when a controller
> fails over?

Typically, you would use the configuration software on the array combined
with a array specific driver on the host to map a particular instance of the
LUN so that only one is recognised on the host. It's not clear from the post
who the array vendor is, NEC? If NEC, than what model, and have you talked
to your NEC support folks?


--
Nik Simpson

[ Post a follow-up to this message ]

Nik,

If we can mask the 2nd LUN on the HBA, then is it still possible for the LUN
to failover to the 2nd controller and the mapping on the HBA to remain
unchanged?

For reference it's an S1300 unit from NEC, we have spoken to them and they
believe it's an issue that we should resolve with our HBA drivers.

"Nik Simpson" <n_simpson@bellsouth.net> wrote in message
news:KEGyc.28177$Kd5.17790@bignews4.bellsouth.net...
> dwayne wrote: 
>
> All things being equal, with nothing done to prevent it, this would be
> normal behaviour, i.e. both controllers present the LUN and when the HBA
> initializes it sees two instances of the LUN.
> 
>
> Typically, you would use the configuration software on the array combined
> with a array specific driver on the host to map a particular instance of
the
> LUN so that only one is recognised on the host. It's not clear from the
post
> who the array vendor is, NEC? If NEC, than what model, and have you talked
> to your NEC support folks?
>
>
> --
> Nik Simpson
>
>

[ Post a follow-up to this message ]

dwayne wrote:
> Nik,
>
> If we can mask the 2nd LUN on the HBA, then is it still possible for
> the LUN to failover to the 2nd controller and the mapping on the HBA
> to remain unchanged?

You need software on the host that is aware that the two lUNs are actually
the same physical LUN being presented via two different paths. The Microsoft
MPIO extension to W2K & W2K03 should be able to do this, but you'll need to
get it from Emulex.

>
> For reference it's an S1300 unit from NEC, we have spoken to them and
> they believe it's an issue that we should resolve with our HBA
> drivers.

Sounds like they are right, I'd start by contacting EMULEX tech support,
they may have MPIO or their own equivalent to handle this situation. That
said, I'm surprised that NEC don't have the ability to do LUN masking and
failover within the array and hide it completely from the host.


--
Nik Simpson

[ Post a follow-up to this message ]

"dwayne" <dwaynev@wn.com.au> wrote in message
news:40cafa7f@quokka.wn.com.au...
> We've just installed a new FC SAN and I've just got a couple of questions
> about some issues we're seeing.
>
> Basically we've got single ported dual controller's connected to 2
switches.
> Each host has a single HBA connected to a switch.  The switch is fabric
> enabled and both are interconnected.
>
> Our issue is that our Windows servers (2k & 2k3), can see duplicates of
> their own storage.  i.e.: we have created a LUN for host1 that's 300 GB in
> size.  Host 1 can see 2 instances of this storage and is able to write to
> both instances of storage.
>
> We think this is due to both controllers advertising the presence of the
LUN
> and our host's HBA being permitted to access it, is seeing it as 2
different
> LUN's.  Is this likely to be the cause, or is it possibly something to do
> with the cross-connect in the switches?
>
> Is there a way to perhaps force the HBA to only see 1 instance of the
> storage, but still allow the LUN to be available when a controller fails
> over?
>
> Should mention that the san is an NEC, the switches Brocade Silkworm's and
> the HBA's are Emulex's.
>
> Any help would be appreciated.
>
> Dwayne.
>
You need  multipathing  software, EMC has PowerPath, HP uses SecurePath.
Each controller sees a  LUN, si and thinks it has unique LUNS on each
channel. Don't know what NEC uses for multipathing software. Other options
are to remove one of the HBAs, remove one of the fibre cables, or only zone
one path.

[ Post a follow-up to this message ]

On Sat, 12 Jun 2004 20:47:17 +0800, "dwayne" <dwaynev@wn.com.au>
wrote:

>We've just installed a new FC SAN and I've just got a couple of questions
>about some issues we're seeing.
>
>Basically we've got single ported dual controller's connected to 2 switches
.
>Each host has a single HBA connected to a switch.  The switch is fabric
>enabled and both are interconnected.
>
>Our issue is that our Windows servers (2k & 2k3), can see duplicates of
>their own storage.  i.e.: we have created a LUN for host1 that's 300 GB in
>size.  Host 1 can see 2 instances of this storage and is able to write to
>both instances of storage.
>
>...........................................

Dwayne,

Didn't  your sales guy or the sales engineer or the NEC SAN architect
ever mention multipathing prior to the sale????  If not, I'd be vary
wary about the SAN config you have and how it's going to perform.

Sounds like you are new to SAN, not that there's anything wrong with
that but so that you know, multipathing is like one of the wheels on a
car and multipathing is a part of a SAN 101 education.  It is a
critical part of a SAN design.  I can't believe NEC didn't provide you
with a proper solution.  That says a lot about NEC and their SAN
solution.

BTW, from my personal experience MPIO was very unreliable and I
wouldn't recommend it.  EMC has Powerpath but you can't use it because
you don't have have a Symmetrix or a Clariion.  You may want to look
at Veritas DMP.   Powerpath and DMP are very good products.

Boll Weevil

[ Post a follow-up to this message ]

Boll Weevil <dryer@maytag.com> wrote in message news:<ahg8e0t5fcb71ouvg6llmh9hrakhk0nmal@4ax
.com>...
> On Sat, 12 Jun 2004 20:47:17 +0800, "dwayne" <dwaynev@wn.com.au>
> wrote:
> 
>
> Dwayne,
>
> Didn't  your sales guy or the sales engineer or the NEC SAN architect
> ever mention multipathing prior to the sale????  If not, I'd be vary
> wary about the SAN config you have and how it's going to perform.
>
> Sounds like you are new to SAN, not that there's anything wrong with
> that but so that you know, multipathing is like one of the wheels on a
> car and multipathing is a part of a SAN 101 education.  It is a
> critical part of a SAN design.  I can't believe NEC didn't provide you
> with a proper solution.  That says a lot about NEC and their SAN
> solution.
>
> BTW, from my personal experience MPIO was very unreliable and I
> wouldn't recommend it.  EMC has Powerpath but you can't use it because
> you don't have have a Symmetrix or a Clariion.  You may want to look
> at Veritas DMP.   Powerpath and DMP are very good products.
>
> Boll Weevil

Good advice, however, I'd like to make one observation:
Powerpath supports "Hitachi Data Systems (HDS) Lightning, HP XP
(Hitachi OEM), and IBM Enterprise Storage Server (Shark) storage
systems."

From: http://www.emc.com/products/software/powerpath.jsp

And from my own experience - powerpath on the HP XP line works quite
well.

[ Post a follow-up to this message ]