looking for further documentation covering nw6.5 buildin multipath support.
Rudolf Thilo
I have to setup a NW6.5 server using this new hardware:
IBM DS4700 FC
IBM BladeCenter H
IBM HS21-XM
Emulex 1105 FC HBA
The Test System's startup.ncf looks like this:
------------- startup.ncf ---------------------
set multi-path support = on
LOAD ACPIDRV.PSM
######## End PSM Drivers ########
LOAD SCSIHD.CDM /AEN /DEBUG
######## End CDM Drivers ########
LOAD LPFC.HAM SLOT=101 TOPOLOGY=2 /LUN RSCNZONE=1 node_timeout=1
LOAD LPFC.HAM SLOT=102 TOPOLOGY=2 /LUN RSCNZONE=1 node_timeout=1
######## End HAM Drivers ########
LOAD KEYB.NLM Germany
LOAD CHARSET.NLM CP850
------------- eol startup.ncf ---------------------
When I comment out the "multi-path support = on, then in nssmu all the devices are
presented double, so I think I can assume, that both path's are "visible" to the
system. Same is reported by the BIOS of the Emulex HBAs when entered at boot time.
What else do I need to do to enable and use the buildin multipathing support of
NW6.5.8?
I've just found 2 topics in the documentation:
1.) os_util_nw.pdf, page 137, Multipath Connection Failover Commands (NetWare):
All these commands return values suggesting that currently multipathing isn't
available.
2.) stor_nss_lx_nw.pdf: After adding the ...
/allpaths /portnames
... to the lpfc load lines I rebooted that server. At that next reboot the server
hangs after loading the 2nd instance of the LPFC.HAM, ("A scheduled work to do took
over a minute ...."), after another reboot the server boot terminates at the very
beginning with a message like "unable to read c:\nwserver\server.exe". The
server.exe is binary identical with that one of another functional server.
So I'd like to read up more detailed documentation covering the buildin multipath
support of NetWare 6.5(sp8)
Reading up a lot of postings here it looks like multipathing isn't a too complicated
thing -- well, I fail so far :(
Who could pls point me out where to find more information?
Thanks, Rudi.
Rudolf Thilo
>> 2.) stor_nss_lx_nw.pdf: After adding the ...
>>
>> /allpaths /portnames
>>
>> .... to the lpfc load lines
>>
>
>Those aren't parameters the lpfc driver will recognize.
I had seen, that all example load lines for LPFC.HAM didn't have these params.
As I found them in the NetWare Docs, I thought they might be generally valid syntax
for all HAMs multipathing capable...
>Hmmm. You're booting from the Fibre Channel?
Yes.
>Is the "unable to read"
>message from DOS or from Netware?
The screen already switched to the NetWare typical layout. From the timing it's
prior the line "Press Esc to ..." would show up.
>I no longer have access to a multipath system, but it all seemed pretty
>simple to me. Keep the set command in startup.cfg, and do a "list
>failover devices" and let us see the output.
Noting is listed, when entering "list failover devices".
When I chance startup.ncf to "set multi-path support = off", then nssmu will display
the HDD devices double.
>The lpfc driver rarely needs any non-default parameters; the
>node_timeout=1 is about the only one I'd recommend and you already put
>that in. There are a lot more parameters, of course, but most of them
>are for rather special cases.
Still wondering what I might mix up here.
Somewhat frustrated, Rudi.
Rudolf Thilo
>Those aren't parameters the lpfc driver will recognize.
Does this startup.ncf file looks like a valid one to make use of NetWare's buildin
multi-path support?
--------------- startup.ncf ------------------
set multi-path support = on
LOAD ACPIDRV.PSM
######## End PSM Drivers ########
LOAD SCSIHD.CDM /AEN /DEBUG
######## End CDM Drivers ########
LOAD LPFC.HAM SLOT=101 TOPOLOGY=2 /LUN RSCNZONE=1
LOAD LPFC.HAM SLOT=102 TOPOLOGY=2 /LUN RSCNZONE=1
LOAD LSIMPTNW.HAM SLOT=10025
######## End HAM Drivers ########
LOAD KEYB.NLM Germany
LOAD CHARSET.NLM CP850
------------- EOF startup.ncf ----------------
I'm using the LPFC.HAM driver from the IBM -> Emulex certified driver site, which is
version
LPFC.HAM
Loaded from [C:\NWSERVER\DRIVERS\] on 16 Nov 2009 17.23.05
(Address Space = OS)
NWPA Emulex Light Pulse HAM driver
Version 2.21.04 10 August 2006
Copyright 2004-2006 Emulex Corp. All rights reserved.
That's an older one than that one included in NW6.5SP8:
Version 2.22.01 Montag, 16. Juli 2007
Copyright 2004-2007 Emulex Corp. All rights reserved. Collective work.
NWPA Emulex Light Pulse HAM driver
The reason is, that the 2.22.01 one caused a "A scheduled work to do took over 1
minute ..." msg after the 2nd. load of LPFC.HAM. So I checked the IBM site, which is
pointing to a Emulex site, and verified with the IBM Hotline, that this version
2.21.04 is the certified one that we shall use *ONLY*.
Regards, Rudi.
Rudolf Thilo
>
>(re: logfile posted)
>
>Strange -- the messages just stop about the time things are to be
>handed over to NSS. And the swap file is getting set up.
>
>If you pull one cable, then it boots properly?
This is a *BLADE* server, so I cannot pull a cable. I just can disable one of the FC
ports at the Brocade switch. When I do so with that one "pointing" towards the 2nd
FC port of the blade, yes, then it's booting fine.
>This is the
>one-path-failed version of multipathing ;-}
Sorry, I miss the meaning of that sentence.
>Also, is it possible to try
>it without loading LSIMPTNW.HAM?
No difference.
>At a first reading, I don't see anything in the logfile that would clue
>me in. Everything SCSIHD.CDM is doing seems very normal. But I'll take
>another look soon.
Thanks for all of your replies!
Regards, Rudi.
Rudolf Thilo
>One more test of that configuration: disable the -other- port and make
>sure that works. If it does, then I think that pushes the failure up
>into scsihd.cdm or mm.nlm, and you'll have to ask the kind people in
>Provo why that happens. If it doesn't work, then I'd suspect a hardware
>component, but to find which one you might need to swap the ports in the
>Brocade switch.
I'm with you, that this all might be a hardware issue, so I'll check primarly in
thea direction.
Thanks so far, I'll report back or ask further questions, when the checking of the
HW is done.
Regards, Rudi.
So this is the story now:
Meanwhile whe had twice a strange situation, first of all I need to describe the
full environment:
Description:
---------------------------
IT-Room 1:
BladeCenter H: "BC1"
Brocade 4Gb FC Switches, "BC1-FCSW1" and "BC1-FCSW2"
DS4700, the productive storage system for all blades, 2 Expansion Units: "DS4700-1"
IT-Room 2:
Main Hardware identical to room #1.
"BC2"
Brocade Switches "BC2-FCSW1" and "BC2-FCSW2"
"DS4700-2", mirror storage system.
The two switches #1 of both blade centers are fabric-1
the two switches #2 are fabric-2
As still 1 port of all switches is unused, there was no need to add external
switches, the 4 internal brocade switches in the BladeCenters are the only one we
use.
Trouble situation:
---------------------------
At Friday evening, after hours, so 100% sure no massive traffic LAN / FC, all the
blades in BC2 lost their LUNs located at DS4700-1.
At this situation the web interface of the FC switches presented:
BC1-FCSW1 did *NOT* "see" BC1-FCSW2, his fabric partner, but at the very same time
the web interface of BC2-FCSW1 *DID* "see" BC1-FCSW1. Error Counters on all switch
ports did *NOT* increment.
The IBM Support worked for hours on the supportshow data of all 4 switches.
Rebooting the BC2-FCSW1 using the management module of the BladeCenter (AMM) didn't
help out, it was requiered to unplug power of BC2, and everything was fine again.
This was done on Monday early in the morning.
The exact same situation showed up on Tuesday evening, at that time not a single
person was in the office, no backup jobs or whatsoever were active. Same procedure,
hours and hours of talking to IBM, finally BC2 had to be power cycled again. On
Wednesday the SFPs for the interswitch link were replaced (that's port #18 for both
of them). Afterwards all the counters were reset, I just pulled another supportshow,
DS4700 Support Data Collection, AMM Support Data. I had a quick look into the files,
couldn't see something unusual, so I'm waiting what IBM might check out from the
data.
>Good luck!
>
>/dps
Regards, Rudi.
--
IT-Beratung Rudolf Thilo
Schweinfurter Str. 131
97464 Niederwerrn
Germany
Rudolf Thilo
>In the abends forum, Marcel had this in the FATFS thread:
>
>> "Regarding your issue, are you using multipathing by any chance? As far
>> as I know, combining booting over the SAN with multipathing is
>> problematic.
Thanks for referencing Marcel's posting. I asked another question over there,
regards, Rudi.