Problem adding new HSG80s to SAN

[Rob Buxton]

We have a SAN consisting of 2 x Fabric Switches (v3.0.2f). Attached to this are: 4 x WNT (NT4 SP6)Servers, 2 x HSG80s (ACS 8.7F) and a Storagetek L80 Tape Library with SPM (3.1A SP1) All works well.
I've now tried to attach 2 new HSG80s. The new HSGs recognised the connections from the Fabric. I renamed connections, created RaidSets / Units and enabled access to a Server for the Units. But the Server does NOT see the new raidsets. Neither SPM or SWCC sees the Controllers or devices. Unit numbers are Unique. Each Server has dual FCAs, All HSGs configured for Multibus Failover , SCSI-2 with NoCCL.
From LPUTILNT, under Persistent Binding, the IDs for the two original HSGs are 0 & 1, for the new HSGs they're 2 & 4 on one card and 3 & 5 on the second. LPUTILNT shows Automap Enabled.
This is the only difference I can see between the two set ups.
Do the IDs need to match before SPM recognises the device?
The Storagetek Tape Library only has a single connection to the SAN, and shows up as ID 2 on one of the FCAs.

[Vincent Fleming]

If I remember correctly, the HSG80 has Selective Storage Presentation... I would check to be sure you have allowed the hosts to see the LUNs with SSP.

[Spike Burkhardt]

Rob,

First off, let me say that I barely know HPUX. I am an itermediate level SAN guy. Also, I'm assuming that the problems are on the HPUX box, not the NT boxes.

I think you're going to need to reboot or do a ioscan -fnC disk command. When you talk about SPM, are you talking about Secure Path?

[Eugeny Brychkov]

Rob, do you have special zoning set up in the switch? Could you please attach supportshow?
Eugeny

[Rob Buxton]

Vincent,
To allow the NT boxes to see the LUNs you do an enable_access_path on the HSGs for the server connections required. This has been done.

Spike,
There is NO HPUX involved, it's all NT. And yes, SPM is Secure Path Manager.

Eugeny,
There is NO Zoning on either Fabric Switch. We did nothing on either fabric switch to get the first two HSG80s to be seen by the Servers.
Do you still want to see the supportshow output?

[Eugeny Brychkov]

Yes, please, from both brocades
Eugeny

[Rob Buxton]

Eugeny,

supportshow output included, both switches in the one attachment.

Ports 0 & 1 on each switch are the original HSGs - that work
Ports 7 & 8 on each switch are the new HSGs.
Port 6 on the bottom switch is Storagetek tape library.

And, thanks to all for the interest!

[Rob Buxton]

Eugeny,

supportshow output included, both switches in the one attachment.

Ports 0 & 1 on each switch are the original HSGs - that work
Ports 7 & 8 on each switch are the new HSGs.
Port 6 on the bottom switch is Storagetek tape library.

And, thanks to all for the interest!

[Eugeny Brychkov · Accepted solution]

I see only supportshow for bottom switch in attachment. Please attach both.
I see correct set-up of SAN from the switch, but there are excessive errors on all ports:

pt0: 87713133 frTx 347197482 frRx 5863893 LLI_errs.
pt1: 382312157 frTx 1036330400 frRx 5587594 LLI_errs.
pt2: 1048376421 frTx 1157645895 frRx 4045412 LLI_errs.
pt3: 107916250 frTx 3030617740 frRx 3546477 LLI_errs.
pt4: 376995610 frTx 1924256226 frRx 3299988 LLI_errs.
pt5: 8048289 frTx 513797466 frRx 6539681 LLI_errs.
pt6: 1861240540 frTx 157683787 frRx 5171770 LLI_errs.
pt7: 55600 frTx 86402 frRx 18657151 LLI_errs.
pt8: 28864 frTx 46211 frRx 24652653 LLI_errs.


Ports 7 and 8, comparing to frames passed have huge number of errors. In addition there're events on them (and on port 4 - LP8000):

port 8: 1 p8sem max:500 tReceive Id 0x10303300 MUTEX owner: none queued_tasks: none
tReceive was blocked for 500 ms by tFcph (120) at 12:51:29.549
_vxTaskEntry + 0x10 (0x10731174)
_fcphTask + 0x34 (0x10564f94)
_portPoll + 0x104 (0x10469464)
_bloomPoll + 0x8bc (0x1049e08c)
_bloomAttach + 0x9f0 (0x10494cb0)
_bloomSpeedInit + 0 (0x104a3330)

Unfortunately I can not decode this event :o(

What can I propose right now? Check FC cables are of correct types; Check if NT's device manager sees newly created LUNs. I'm not familiar with HSG80s, but looking to your phrase "created RaidSets / Units and enabled access to a Server for the Units" I would ask you to check that you enabled HSG80s controllers access for correct server HBAs (remember, you have one HSG controller connected to one HBA and another controller to another HBA).
About supportshow errors: try operating your SAN and see how fast errors are accumulating
Eugeny

[Rob Buxton]

Eugeny,

Many thanks, I suspect the large number of errors on ports 7 & 8 are due to the fact I've rebooted the HSGs a lot while trying to sort this out.
Cables are as supplied by HP, they're the same type / spec as all the others.

Listing for top switch attached.
For a while the HSGs were on ports 6 & 7, I moved them along one as part of a test.

[Eugeny Brychkov]

Supportshow is clean, except same events "tFcph was blocked for 633 ms by tReceive (85) at 12:26:17.366".
I feel that your issue is HSG80s config problem. As soon as you create LUN in array, it appears for NT it should catch it and display "New hardware found...".
Do you need to create some specific LUN (like LUN 0)? Do you need to force to set HSG up to work in N-port (not to use port-type-discovery)? Do you need to map anything? Do you need to check security/LUN mapping tables? etc.
Good luck!
Eugeny

[Rob Buxton]

Eugeny,

Thanks, I'm not convinced it's the HSG because I've compared it to the ones that do work.
The main difference is the difference in IDs as reported from the Utility on the NT Server.
The HSGs can certainly see the connections from the Server, and the NT Server utility seems to be able to see the HSGs.
There's no LUN 0 on either set of HSGs, and it's not recommended if you're planning on going to SCSI-3 on the HSG80.
I'm getting HP in tomorrow, I'll post back anything discovered then.

[Spike Burkhardt]

Rob & Eugeny,

I've been told by our TAM that NT won't see the new unit until the system is rebooted. Please let me know if this is wrong!

Spike

[Tore_2]

Hi ROB
1, verify that connection type = winnt,.
2, try implementing three zones, to isolate the problems.

Zone1 (backup) = serverHostprots and backup library port.
Zone2 (storage1) = "old" HSG80ports and serverHostports.
Zone3 (storage2) = New HSG80ports and ServerHostports.

(You only need two zones in the fabric with only storage)

And Spike in Window NT you can scan the SCSI bus with the search for new tape device button in control panel, this will detect new disks as well :) without rebooting.
/Tore

[Tore_2]

Hi ROB, again
I???m not sure, but I think I have read that the maximum numbers of active controller ports/HBA is 4 for Windows NT, these numbers keep changing all the time, ask HP when they are at your site..

[Peter O'Connor]

Just a thought... What's your unit No.? If it's over 7 then you'll have to enable large luns to see it on the NT box. (Talk to HP if so).

Else are you sure you havn't run into other constraints.. i.e. max connections?

[David Devlin]

As Peter suggested, check that you have large lun support installed and also take note of what luns you are using. For ACS ver 8.6 you can use 0-63, I don't know how many you can have for ver 8.7, I would suspect the same!

[Rob Buxton]

Hi All,

Thanks for the responses.
We've had HP in who are now going to try and replicate the problem on their SAN.
We still believe it may be because we've got a conflict of IDs at the Server end.
So, at this point we believe it to be a Server issue, possibly related to Persistent Bindings on the FCA Card. We believe the fact it sees the HSGs as 2 & 4 on one card and 3 & 5 on the other may be causing problems. I'll let you know when we've done further testing.

[Rob Buxton]

Hi All,
Disabling Automap and enabling Persistent Binding on the Server's FCA allowed Civic to see the Disks. HP are reviewing what we did and why it worked!
Thanks for all the input.