Probably K-Class immediate SCSI bus termination problem

[Ralph Grothe]

Hi K-Class admins or techies,

we have an old left over K-Class (for which there is no support anymore) which is causing me some trouble.

I was charged with the task of installing a current OS and some software on it, and configuring it for server duties in one of our LANs.

But at the moment I can't even take the first step since that damned thing (i.e. the service processors routines) won't find any hard disks to install on.
It only reports the CDROM, DAT, and LAN device as bootable candidates.

Unfortunately this box has no internal disks at all.
But in the rack are two external storage enclosures of make A3312A.

The K-class box itself is furnished with an I/O expansion slot that hosts a 4 port HP-HSC I/O card.

I downloaded, printed, and carefully read all the relevant documents from here:

http://www.docs.hp.com/hpux/hw/index.html#K-Class%20Server

The HP-HSC Expansion I/O Installation Guide states that the HW paths assigned to this board are (from bottom to top) 8/0, 8/4, 8/8, 8/12.

I tried all sorts of connections (i.e. HBA different sockets on HP-HSC card to sockets on storage enclosure, using differential cabling and terminators (the disks are also DF of course)

I'm sure to have connected all correctly as I clinged to the figures and tables from the A3312A Installation/Service Manual regarding cabling and SCSI ID settings.
I even dismounted the whole enclosure from the rack and manually set the SCSI IDs of the slots at their respective dip switch (according to the remarks in the manual, viz. not to assign ID 0 to either slots A3,4 or B3,4 (counted from bottom to top starting with A|B1).
I'm also convinced to not have reused IDs, but the switches' positions are *very* difficult to realize.
At this point I have to utter that the dip switches really are crap because one cannot see the sitches' positions while the enclosure is mounted in the rack, even if one crawls with a torch into the rack.
It would have been a great help if HP engineers would have at least given the switches a kind of reddish colour, or better had made them bigger.

All these attempts were in vain though.
I now reduced the SCSI chain to the least length, i.e. only 4 disks in one slot (either A or B).

I also carefully looked at the pins of the male SCSI plugs for bent or broken pins.
With regard to this point of failure I think that the Centronix kind of plugs are a much better design.

Despite my efforts the service processor won't find devices under any HW path 8 when issuing a SEA (or IN IO) command.

What puzzles me is that the LEDs of the HP-HSC I/O card at the back next to each scsci socket are lid, where the label reads a "term" which I interpret as termination.

I would think the sockets are terminated internally as long as no socket is plugged in.
But I also gave it a try to plug DF terminators on each unused socket, which didn't help either.

Has anyone with a K-class and external disk storage encountered similar difficulties who can give me any hint what else to try, or tell me what I'm doing wrong?

Rgds
Ralph

[Alexander M. Ermes]

Hi Ralph.
on the back of the enclosures you should put active taerminators on the open connectors.
We use the Jamaica box for sample and we had to go this way. Otherwise the SCSI bus i not properly terminated.
Rgds
Alexander M. Ermes
p.s. hope i could help

[V.Tamilvanan]

Hi,
Normally this sort of problem occurs either due to SCSI ID clash or terminstion.

Try by terminate all the cards SCSI interface.
Change the id of HSC Scsi card by changing the dip switches on the card and check for the internal terminator which is on the cards.

HTH

[Ralph Grothe]

Alexander, Tamil,

that's interesting.
I didn't find any note that active termination on the card's sockets is required in my docs.
I will check this tonight while I have to access our server bunker for maintenance anyway.

Btw, please excuse my foolish question,
how can I identify by looking at the terminator if it's an active one?
Maybe to give you a rough description, it is shaped like a parallelogram (i.e. it has some taper to its end)

[Stefan Farrelly]

Hi Ralph,

Tha active terminators normally have a light on them so when plugged in they light up to let you know theyre terminating - but I digress - you dont need terminators at all. We dont use them on ours.

I would remove all disks bar 1 - I suspect one of the disks you have is broadcasting SCSI errors or resets which cause your K to not see any. Or its possible the unit itself has a faulty scsi controller causing the same problem (ive seen this happen - can be caused by a dip switch for a scsi setting being broken off for example). Can you test the disk unit on another server ? any L or N with a DF 68pin controller can connect up to it.

[Ralph Grothe]

Stefan,

so I thought, that no active termination is required on the card's sockets itself.
I'm sure it would have been mentioned explicitly in my docs if this wasn't the case.
Digressing as well - no, there are no LEDs in the terminators.
Thus I guess mine are only passive.

Good suggestion,
we still have an unused D-Class where I should connect my disk enclosures to for verifying that it's neither the enclosures nor any disk in them.

[Alexander M. Ermes]

Hi Ralph.
Sorry for a misunderstanding.
Active terminators on the open sockets on the disk enclosure, not on the interface.
Are you sure, that your cables are not too long ?
Rgds
Alexander M. Ermes

[Eugeny Brychkov]

Ralph,
1. about K-class. It can have HVD SCSI adapters in the HSC slots or SE SCSI in HP-PB slot.
2. Jamaica A3312A could be configured in SE or HVD mode. Look at the disks installed in it. Which label do they have? Pink DF or Yello SE? If DF, then connect this channel to HVD, if SE - then to SE adapter.
In your situation, I would remove all disks from the jamaica, disconnected all cables from its back and tried to assemble everyting from the scratch (assume you have DF disks!):
- connect jamaica's channel A to any of server's HVD HBAs (HSC installed);
- attached to another channel A connector HVD terminator (make sure this is correct one - A1658-62024!!);
- install one disk into any slot of channel A. Try searching for this drive from PDC, look at drive's led to understand if server queries this disk or not.
Please note that SEA IPL looks for disks on which OS is already installed (you'll see it telling 'searching for potential boot devices'). Issuing SEA should find all accessibel storage devices
Eugeny

[Ralph Grothe]

Hello responders,

many thanks for your valuable hints.

Yesterday I had a chance to ask an HP Service Engineer, and he gave me the decisive clue to the missing link (oops, unintended pun here?).

First of all, my termination was all ok;
the cables are much shorter than the allowed ca. 25m for FW SCSI;
all the disks are DF and are operating ok;
disk storage unit is also ok.

I verified all this when I connected the unit to the one onboard FW SCSI controller that I've overlooked.
Connected to it the service processor's SEA command merrily reported the "new" disks.

The thing, as the SE explained to me, and what I forgot to convey in my query here, was that the left over K-Class box is only one of the twins that used to form a cluster in a more splendid past.
Thus the I/O expansion board obviously must have been configured to share disks with another box's SCSI controllers.
Since this link had been decapitated (the second K-Class had long been given back to HP) it correctly refused operation.

The SE told me I should dismantle the I/O board and look if the onboard terminators are present.
If not I should look for them at the place where they'd been stored.

Oh my, let's hope I can find this place.

Rgds
Ralph