We have an MSA 1500 with dual controllers. Each had a single HBA, and each
HBA attached to a seperate switch. When one of the HBA's in a server failed,
we discovered that the path to the standby controller did not function,
obviously. So we were told my HP to buy a switch or a hub to rpovide access
from each controller to each switch- the hub was signicantly cheaper.

So now we have 2 - 2 port hubs rather than a single HBA. And when I enable
the port where the second port in the hub is attached to (on the opposite
switch of the working one) we get 2 failed ports. I've contacted HP, and
they're lookign into it... any ideas? Is there some kind of conflicting
path?

[ Post a follow-up to this message ]

I'm not entirely sure I follow you. "Each had a single HBA" Each MSA1500
controller? They're normally not called HBAs, rather I/O modules.
In any case, each controller connected to a seperate switch is perfectly
valid high-availability configuration.

Or do you mean each server had a single HBA? Can't see why you should need
to mix in hubs or more switches.

If you could draw me a simple diagram of each of the setups you've tried I'd
be happy to help. Email adress is valid. Please include server OS in the
diagram
/charles

"Kat" <kit-kat bar> skrev i meddelandet
news:429f3bd3$1@obsidian.gov.bc.ca...
> We have an MSA 1500 with dual controllers. Each had a single HBA, and each
> HBA attached to a seperate switch. When one of the HBA's in a server
> failed,
> we discovered that the path to the standby controller did not function,
> obviously. So we were told my HP to buy a switch or a hub to rpovide
> access
> from each controller to each switch- the hub was signicantly cheaper.
>
> So now we have 2 - 2 port hubs rather than a single HBA. And when I enable
> the port where the second port in the hub is attached to (on the opposite
> switch of the working one) we get 2 failed ports. I've contacted HP, and
> they're lookign into it... any ideas? Is there some kind of conflicting
> path?
>
>

[ Post a follow-up to this message ]

Sorry Charles  Yes, I meant each I/O module on the controllers.

We have 2 switches, and each server/MSA has a connection, one to each
switch. Now, the msa has 4 connections- each controller conencts to both
switches. At least this is what we're trying to do- the hubs don't seems to
like having multiple connections.When I enable the second port on the
switch, that the msa controller is connected, both the ports that that
controller is connected to will fail. If I disablet he second port, it works
fine.

I can draw a quick diagram if you need one.

Thanks

"Charles Morrall" <charles.morrall@telia.com> wrote in message
news:cmIne.139248$dP1.489573@newsc.telia.net...
> I'm not entirely sure I follow you. "Each had a single HBA" Each MSA1500
> controller? They're normally not called HBAs, rather I/O modules.
> In any case, each controller connected to a seperate switch is perfectly
> valid high-availability configuration.
>
> Or do you mean each server had a single HBA? Can't see why you should need
> to mix in hubs or more switches.
>
> If you could draw me a simple diagram of each of the setups you've tried
I'd
> be happy to help. Email adress is valid. Please include server OS in the
> diagram
> /charles

[ Post a follow-up to this message ]

I don't see any reason for the hubs. Infact, they could be making matters
worse.

Each server should have two HBAs, connected to the respective switches. So
should the MSA1500 controllers, each I/O port connected to the respective
switch. The MSA1500 controllers are active/passive, meaning the entire load
is on one controller. The redundant controller is idling, waiting for either
the controller to fail or a path from a server to fail. This means any
single-attached server will lose connection if the controller pair fails
over.

Things to check:
Do all servers have two HBAs, each connected to seperate switches?
Are the HBA drivers up to the minimum supported version?
What is the OS on each server?
Do you have multipath drivers loaded? SecurePath or MPIO? What version?
Is there any zoning on the switches? What model of switch? Firmware on
these?
Is the firmware updated on the MSA1500?
Have you confirmed there are no bad cables, SFPs, HBAs?
How is Selective Storage Presentation (SSP) configured on the MSA1500? Are
there two WWID entries for each server?


"Kat" <kit-kat bar> skrev i meddelandet
news:429f68f8$1@obsidian.gov.bc.ca...
> Sorry Charles  Yes, I meant each I/O module on the controllers.
>
> We have 2 switches, and each server/MSA has a connection, one to each
> switch. Now, the msa has 4 connections- each controller conencts to both
> switches. At least this is what we're trying to do- the hubs don't seems
> to
> like having multiple connections.When I enable the second port on the
> switch, that the msa controller is connected, both the ports that that
> controller is connected to will fail. If I disablet he second port, it
> works
> fine.
>
> I can draw a quick diagram if you need one.
>
> Thanks
>
> "Charles Morrall" <charles.morrall@telia.com> wrote in message
> news:cmIne.139248$dP1.489573@newsc.telia.net... 
> I'd 
>
>

[ Post a follow-up to this message ]

"The redundant controller is idling, waiting for either
the controller to fail or a path from a server to fail. This means any
single-attached server will lose connection if the controller pair fails
over."

This is incorrect- if a single connection to a server fails, it will fail
over to the other switch. But the passive/redundant controller does NOT
become active- it will only become active if the other controller fails. So
essentially, that path is useless. It can connect to the passive controller,
but that controller can't do anything.

As it stands, everything is working, in the way you describe it. We have
multiple servers attached to the switches, all dual path, all with
securepath and W2003, and yes there are two WNNs for each server showing on
the MSA. The config is correct- it's just a matter of getting the redundancy
to work for each controller (attaching each controller to each switch).

Since HP recommended we do this, I'd assumed it would work, but I can't find
anyone else who has it configured this way. I'm starting to have doubts 



--
Kat
MCNGP # 29, MCDBA # ? of Millions

What woud you do for a Kit Kat bar?
"Charles Morrall" <charles.morrall@telia.com> wrote in message
news:dYSne.26293$d5.177010@newsb.telia.net...
> I don't see any reason for the hubs. Infact, they could be making matters
> worse.
>
> Each server should have two HBAs, connected to the respective switches. So
> should the MSA1500 controllers, each I/O port connected to the respective
> switch. The MSA1500 controllers are active/passive, meaning the entire
load
> is on one controller. The redundant controller is idling, waiting for
either
> the controller to fail or a path from a server to fail. This means any
> single-attached server will lose connection if the controller pair fails
> over.
>
> Things to check:
> Do all servers have two HBAs, each connected to seperate switches?
> Are the HBA drivers up to the minimum supported version?
> What is the OS on each server?
> Do you have multipath drivers loaded? SecurePath or MPIO? What version?
> Is there any zoning on the switches? What model of switch? Firmware on
> these?
> Is the firmware updated on the MSA1500?
> Have you confirmed there are no bad cables, SFPs, HBAs?
> How is Selective Storage Presentation (SSP) configured on the MSA1500? Are
> there two WWID entries for each server?
>
>
> "Kat" <kit-kat bar> skrev i meddelandet
> news:429f68f8$1@obsidian.gov.bc.ca... 
MSA1500[vbcol=seagreen] 
perfectly[vbcol=seagreen] 
tried[vbcol=seagreen] 
the[vbcol=seagreen] 
>
>

[ Post a follow-up to this message ]

"Kat" <kit-kat bar> skrev i meddelandet
news:42a08b65$1@obsidian.gov.bc.ca...
> "The redundant controller is idling, waiting for either
> the controller to fail or a path from a server to fail. This means any
> single-attached server will lose connection if the controller pair fails
> over."
>
> This is incorrect- if a single connection to a server fails, it will fail
> over to the other switch. But the passive/redundant controller does NOT
> become active- it will only become active if the other controller fails.
> So
> essentially, that path is useless. It can connect to the passive
> controller,
> but that controller can't do anything.
>

Well, it's supposed to work that way 

Just a thought, are the switches connected to each other?
/charles

[ Post a follow-up to this message ]