Problem Connecting HP Blade to IBM SAN

"ramcke44" wrote in message
...

"Viper" wrote in message
ups.com...

ramcke44 wrote:
"Viper" wrote in message
oups.com...
I was able to get an HP rep out to look at the problem first hand and
after a few days of both of us being completely confused, we turned up
a driver that was hidden on IBM's website which supported connecting
our HBA to an ibm san. after this was installed all is now well with
the blades and the san. I thank you all for your insight.


What type switch are you using, Cisco had a problem with Qlogic blade
server
HBAs.

the fiber switches that we use are a combination of IBM, EMC and HP but
they are all brocade switches.

It seems as if the driver that i said worked, hasn't actually fixed all
our problems so i'm still in search of more things that i can
troubleshoot, if anyone has any new information.

Qlogic blade HBAs use a WWN format that is typically reserved for targets.
Some older switch firmware versions have problems with this as they are
not expecting to zone a target to a target. We ran into this when we first
deployed blades a couple of years ago on Cisco switches. This may not be
the case with Brocade but I would try upgrading the switch firmware first.

World Wide Node and Port Name format is defined by T11 and there was never a
special format for targets. At most, a switch will parse the IEEE address
contained within the WWNN/WWPN to try and determine who the vendor is. Most
WWNN/WWPN are based on the IEEE Extended format, please reference FC-FS
specification, section 14.

What problem are you still experiencing?

Da Moojit

"Moojit" wrote in message
...

"ramcke44" wrote in message
...

"Viper" wrote in message
ups.com...

ramcke44 wrote:
"Viper" wrote in message
oups.com...
I was able to get an HP rep out to look at the problem first hand and
after a few days of both of us being completely confused, we turned
up
a driver that was hidden on IBM's website which supported connecting
our HBA to an ibm san. after this was installed all is now well with
the blades and the san. I thank you all for your insight.


What type switch are you using, Cisco had a problem with Qlogic blade
server
HBAs.

the fiber switches that we use are a combination of IBM, EMC and HP but
they are all brocade switches.

It seems as if the driver that i said worked, hasn't actually fixed all
our problems so i'm still in search of more things that i can
troubleshoot, if anyone has any new information.

Qlogic blade HBAs use a WWN format that is typically reserved for
targets. Some older switch firmware versions have problems with this as
they are not expecting to zone a target to a target. We ran into this
when we first deployed blades a couple of years ago on Cisco switches.
This may not be the case with Brocade but I would try upgrading the
switch firmware first.

World Wide Node and Port Name format is defined by T11 and there was never
a special format for targets. At most, a switch will parse the IEEE
address contained within the WWNN/WWPN to try and determine who the vendor
is. Most WWNN/WWPN are based on the IEEE Extended format, please
reference FC-FS specification, section 14.

What problem are you still experiencing?

Da Moojit
I am not experienceing any problems and I am aware of the IEEE WWN spec, but
in reality targets typically begin with a 5, at least all EMC, HP and IBM
targets on my SAN do. The only exception I have seen is the Qlogic blade
HBAs (there may be others but I have not seen them). According to HP some
switch manufactures employed a shortcut in their OS code to assume that WWN
beginning with a 5 were targets. The result was that the card would login to
the fabric and you could zone the adapter but you would not be able to
present disk. There was a work around to this which we successfully tested
but it was kludgy, the better fix was to upgrade the SAN OS to remove this
obstacle.

"ramcke44" wrote in message
...

"Moojit" wrote in message
...

"ramcke44" wrote in message
...

"Viper" wrote in message
ups.com...

ramcke44 wrote:
"Viper" wrote in message
oups.com...
I was able to get an HP rep out to look at the problem first hand and
after a few days of both of us being completely confused, we turned
up
a driver that was hidden on IBM's website which supported connecting
our HBA to an ibm san. after this was installed all is now well
with
the blades and the san. I thank you all for your insight.


What type switch are you using, Cisco had a problem with Qlogic blade
server
HBAs.

the fiber switches that we use are a combination of IBM, EMC and HP but
they are all brocade switches.

It seems as if the driver that i said worked, hasn't actually fixed all
our problems so i'm still in search of more things that i can
troubleshoot, if anyone has any new information.

Qlogic blade HBAs use a WWN format that is typically reserved for
targets. Some older switch firmware versions have problems with this as
they are not expecting to zone a target to a target. We ran into this
when we first deployed blades a couple of years ago on Cisco switches.
This may not be the case with Brocade but I would try upgrading the
switch firmware first.

World Wide Node and Port Name format is defined by T11 and there was
never a special format for targets. At most, a switch will parse the
IEEE address contained within the WWNN/WWPN to try and determine who the
vendor is. Most WWNN/WWPN are based on the IEEE Extended format, please
reference FC-FS specification, section 14.

What problem are you still experiencing?

Da Moojit
I am not experienceing any problems and I am aware of the IEEE WWN spec,
but in reality targets typically begin with a 5, at least all EMC, HP and
IBM targets on my SAN do. The only exception I have seen is the Qlogic
blade HBAs (there may be others but I have not seen them). According to HP
some switch manufactures employed a shortcut in their OS code to assume
that WWN beginning with a 5 were targets. The result was that the card
would login to the fabric and you could zone the adapter but you would not
be able to present disk. There was a work around to this which we
successfully tested but it was kludgy, the better fix was to upgrade the
SAN OS to remove this obstacle.

HP is wrong. The fact that EMC has 0x5 in it's most signficant byte is
coincidence. A switch (like all nodes) can determine if a node is an
initiator and/or target by inspecting the Initiator bit returned in the PRLI
accept frame. Brocade, McData, Qlogic, etc. switches do not care if a node
is a target or initiator, zoning algrorithm's are not based on this. Sorry
for coming down so hard, but this is the kind of mis-information that get's
thrown around by computer experts all of the time.

Da Moojit

There is a global setting that may have to be changed on the brocade
switch. The hp blade server uses a core switch policy and most brocade
switches not set to that by default. On a brocade to fix this you
must set PortCfgISLMode to 1. Then the ISL will work correctly and
the servers should see the luns.

Where should this ISL mode be set to ON? in my case i have three
brocade switches connected in a chain:
------ ------ ------
| |------| |-------| |
| | | | | |
------ ------- ------

They're all part of the same fabric. should isl be turned on for all
the ports connecting the switches together? or just for the port the
blade actually connects to?

"Viper" wrote in message
ups.com...
Where should this ISL mode be set to ON? in my case i have three
brocade switches connected in a chain:
------ ------ ------
| |------| |-------| |
| | | | | |
------ ------- ------

They're all part of the same fabric. should isl be turned on for all
the ports connecting the switches together? or just for the port the
blade actually connects to?

You should not have to configure anything on the ports they will ISL
automatically. Each switch will needs it's own Domain ID and you will have
to configure this manually on each switch. For each of the three switches,
set the domain ID's 1,2 and 3.

Da Moojit

You need to do the seeting below:

There is a global setting that may have to be changed on the brocade
switch. The hp blade server uses a core switch policy and most brocade

switches not set to that by default. On a brocade to fix this you
must set PortCfgISLMode to 1. Then the ISL will work correctly and
the servers should see the luns.

on the switch as a global variable. The qlogic blade based servers use
a CORE setting on the HP solution. The switch you run uses native and
needs to be changed to core for correct communication. I have
implemented a few of these solutions now and that is always the gotcha.

The answer to this problem was finally found. The answer that seems to
have solved this problem was to use an Emulex HBA instead of the
Qlogic. After installing drivers and checking bios versions for the
Emulex, re-zoning and assigning the LUN, a second hard drive magically
appeared when rescaning for new disks (something that the qlogic card
couldn't do).
So it would seem that the problem was with qlogic and its coexistance
in an hp blade when connecting to an IBM SAN. Thanks for all the
assistance, your insight offered solutions that I would not have
thought of.