Conference mvblab::alphaserver_4100

    Interesting. Both the 2100 and 4000/4100 use an NCR810 as the embedded
    CD/ROM controller.

Check the termination on the internal SCSI bus......

If it is a rackmount it should be at the end of the SCSI cable tucked below the
fan tray. If it is a pedestal the terminator should have been moved to the top
tray. If a CD or tapr is on top make sure the bus has not been double
terminated. 

	Dennis

I have installed 2* RM and 1* pedestal recently.
On boths RMs I had symptoms of .0, pedestal worked fine.
I''ll chech the terminators.
Regards,
	Peter

    	Interesting note,
    
    	I have had problems with the internal scsi and T/S it down to the
    lack of terminator. I look in QRL/Comet/Service Manual/ProSIC IPB, and
    nowhere was there a part number for that terminator that I could find.
    Now I did talk to someone in SBU and I the why I understood it any
    active male term should work, I finally ended up with using a
    Celebris internal active terminator from cable assy 17-03459-02, or
    Term PN# 12-37791-02 (Which is listed by itself for a PB30 DEC 2000
    Model 500. The system in question has ran without any Customer
    compliants so far.

    I checked and the terminator is installed at the end of the bus.
    
    It is part # 12-37791-01.  Are there any other jumpers I need to look
    at.  I would think it was a bad part but I am getting the same
    results on three rm boxes.
    
    Howard

    Hi,
    
    I have the same problem. Using UNIX i have to send "file"comands to the
    tape (TLZ09) if i want to fix the slow down problem of the RRD45. I was
    thinking i have a bad CDROM, but now im not sure that. A NCR driver ?
    
    Any clue?
    
    Tks,
    
    Leandro.

    I have two rm's that demonstrate the exact symptom you described. SCSi
    time outs & aborts. both active & passive termination have been tried (
    active made response worse). I have taken the pedestal sub-system (top
    tray) with 4mm tape which functions fine on pedestal. installed on rm
    and still have problem. CD works on pedestal with no errors. My next
    step is to try SCSI cable and/or saddle module. from known good
    pedestal to rack mount.  Can you say "escalation"
    Todd Murphy
    Dallas MCS
    
    

    I tried swapping some parts from a known good pedestal 4100 into one of
    my rackmounts. I tried in this order, replacing the following parts;
    Internal SCSI cable (with passive term.),, adding 4mm tape with trm
    pwr ena.,term dis., and no term on cable (a good config on the
    pedestal),, PCI mother bd. (saddle module) with original cable and
    passive term, no tape,,  Known good saddle module and horse module with
    original cable, term. All of these attempts yeilded no positive
    results. Under unix 3.2g I continue to recieve excessive load times on
    the cd and continued SCSI aborts and Timeouts entered in decevent.
    Could this be a firmware issue or maybe CPU mother bd.?? I could use
    some input on this one.
    
    Todd Murphy
    Dallas MCS
    

    Hi all,
    I have checked the termination on 2 RMs (.3), all looks fine, but I
    still get aborts on the bus. Is there a solution for this problem ?
    Was it already escalated ? 
    Regards,
    		Peter
    

    We found a fix to our problem.  At the console check 
    pka0_fast if it is set to 1 try setting it to 0 reinit the machine
    
    Once booted try load something and see if it improves.
    
    Howard

    It appears there are several data points here; please correct me where
    I've mis-interpreted or mis-represented the facts:
    
     1. Multiple 4100/4000 installations are seeing slow access times
        on devices hanging off the embedded NCR810. Typically CD/ROM,
        but sometimes configured with tape.
    
     2. Problem is only seen on rackmount systems; pedestals appear to
        be immune to the problem.
    
     3. Slow load times are presumably attributable to SCSI bus errors
        and excessive retries, presumably due to poor signal integrity.
        Multiple machines seen with lots of errors in operating system
        error log. How are the tape drives ordered/configured? At Stage
        II MFG or are they ordered separately and installed onsite?
    
     4. Setting PKA0_FAST to 0 ("slow" mode; default is 1) remedied the
        problem on at least three of the rackmount configs, presumably
        reducing bus signal integrity problems.
    
     5. At least one site found that adding a terminator resolved the
        problem (with PKA0_FAST still set to 1).
    
    Is the above correct? I'll check to see what MFG does re: default
    installation of termination. Presumably either cable length or
    termination is at play here re: the difference between rackmount and
    pedestal. I'll most likely have MFG begin setting PKA0_FAST=0 as the
    default console EV setting. PKA0 is always the embedded NCR810 device.

    The only difference I see between the two versions of 4100's is that
    the rackmount version will only have a CDrom, (no room for a tape
    device).  The pedestal uses an additional cable to go to the tape unit
    if present.  
    
    This morning I upgraded my firmware to SRM 4.8-6 and AlphaBios 5.2-8.
    After loading the same software I didn't see any errors.  So this could
    also be a firmware issue.
    
    I know this doesn't help much.
    
    Howard

    Are you sure this wasn't because you had already set pka0_fast to 0 on
    these machines? This EV is "sticky". Copy kept in NCR810 NVRAM..
    
    The default setting for SRM's PKA0_FAST has always been 1; at least
    since the V1.2-4 console release. I just verified this from the saved
    source files..
    

    Design/support engineering is investigating this as a potential signal
    integrity problem. Until the problem is better understood, the
    recommendation is that customers try setting PKA0_FAST to 0 to see
    whether this helps.
    
     P00>>> set pka0_fast 0

I know of one issue with the OpenVMS 810 SCSI driver, which is documented 
in ALPHASTATION, note 1776.  So, far this problem has only been known to 
occur with OpenVMS V7.1 when a non-TCQ, such as RX26F, is present with the 
CD-ROM.  .6 notes that the problem has been seen under Unix, so this does 
not appear to be related to the known OpenVMS driver issue with RZ26F.


							-Paul

    Hi,
    
    Now i found problems in this bus under VMS, but the slow device is a
    TLZ09. I have changed the dat ( tlz09 ), check terminators, set PKA0_
    FAST to "0" and nothing. Now i'm trying to update the TLZ09 firmware
    to 0167 . I will do that this week. 
    
    4100 5/300
    openvms 6.2-1h3
    firmware from cd 3.8
    
    I would like to now what does the parameter PKA0_DISCONNECT do?
    
    The symptom from DAT is a very slow writing process. If you compare
    with a DAT TLZ07-VA working in the bus of the KZPDA ( ISP1020 ).
    
    Any help?
    
    Tks,
    
    Leandro.
    
    MCS BRASILIA-BRASIL

    Just to make sure..
    
    You have the TLZ09 hanging off the embedded NCR810, yes? Or is it
    hanging off another SCSI controller? It may not be obvious, but
    PKA0_XXXX EVs only apply to the "A" SCSI controller. PKB0_XXXX would
    apply to the next SCSI controller, etc..

    Embedded NCR810. It's the bus of CD (RRD45). DKA500 and MKA600.
    
    PKB0 is a KZPDA (ISP1020) option.
    
    Leandro.

    I have placed a couple of entries early on in this note. My issue with
    a rackmount (2) 4100 still exists. since my last entry I have tried
    horse, saddle, SCSI cable, active/passive term., matching the pedestal
    config (i.e. with TLZ09), setting pka0_fast 0, no combination has had
    success with the acception of a kzpaa with internal scsi connection.
    that config ran flawlessly although this bus then became scsi bus 3. A
    question was raised in on of the previous notes "is this really being
    escalated?" the answer is yes. I have been working with engineering for
    several weeks and will uodate notes as appropriate.
    Todd Murphy
    Dallas MCS
    
    

    I know some of the protos had the RRD45 with terminators installed when
    they shouldn't have been.  You've probably checked the termination
    already, but make sure that you're not terminated in the middle of the
    bus as well as the ends.
    
    Todd

    hi,
    
    My cases .6 and .16 (RRD45 and TLZ09) are on pedestal systems. I dont
    know about rackmount but in my two situations i think i have found the
    problem: TERMINATORS. In both systems the SCSI cable on TLZ09 (frost
    white) dont have terminator and the DAT is not terminated. The CD is
    not terminated too.
    
    Last nigth when installing four new systems (AS4000/pedestal) i've
    checked and saw the difference: the new systems have a different SCSI
    cable on TLZ09. It has a terminator and is better shielded. The PN of
    this cable is : 17-04306-01 (rev.B03). The terminator PN is:
    12-37791-01 rev.A.
    
    In my two systems i have to order two new cables and two new
    terminators cause there is no place to put the terminators the SCSI
    cables are differents.
    
    I think this is the problem i have in both machines. I hope i have
    found the solution.
    
    Tks,
    
    Leandro. 

    
    
    Just a note that TLZ09 (any TLZ drive for that matter) has embedded
    terminator integral to the drive, that can be enabled/disabled via
    jumpers or dip switches.  Bottom line is to make sure that you are 
    terminated on near end and far end and nowhere else in between.
    
    /Frank

    See note 93.46 for a Blitz which may explain the problem seen here..