From:	SMTP%"Info-VAX-Request@Mvb.Saic.Com" 15-SEP-1994 17:34:05.56
To:	EVERHART
CC:	
Subj:	Our VMS 6.1 experiences (FYI)

From: Jim Kirkpatrick <JIMKIRK@UWYO.EDU>
X-Newsgroups: comp.os.vms
Subject: Our VMS 6.1 experiences (FYI)
Message-ID: <01HH55Y9RMPE001AZV@PLAINS.UWYO.EDU>
Date: Thu, 15 Sep 1994 13:43:15 -0600 (MDT)
Organization: Info-Vax<==>Comp.Os.Vms Gateway
X-Gateway-Source-Info: Mailing List
Lines: 162
To: Info-VAX@Mvb.Saic.Com

I thought I'd share our experiences with upgrading our Alpha system
from OpenVMS AXP 1.5 to 6.1.  For the most part, the upgrade went
fairly smoothly, with little user impact.  At the same time, I upgraded
a few products such as the performance monitor (formerly PSDC/PSPA),
and installed new products such as DFO (a.k.a. Defrag, DFG, PolyCenter
File Optimizer).

Note that some of the following may be specific to AXPs, maybe not;
we have no VAXes left.

1.  There is a bug in CONVERT/RECLAIM, discussed here earlier, that
    is quite severe.  Install CSC patch 2062 as soon as the upgrade
    is done, before releasing service or doing much of anything else.

    Of course we upgraded August 1 and the patch became available
    on the 12th.  Thank you, Digital.

    P.S. The VAX version of this patch is 1181.

2.  There is a bug in the virtual I/O cache that bit us several times.
    This is fixed by CSC patch 2045, which is a shadowing patch, but
    seems to have turned into a catch-all patch for many other problems.
    The VIOC fix was also made available after we upgraded, and had
    crashed several times.  Thank you, Digital.

    (by the time we figured out it was a VIOC problem, the patch was
    available; we could have disabled the cache as a workaround if
    anybody at CSC had had an idea there was a VIOC problem in the
    first place).  We found out about the problem, and the patch,
    via Info-VAX, not CSC.

3.  I upgraded the performance data collector, and subsequently had
    several crashes.  We first reported this as a potential hardware
    problem (we were having other problems at the time), and they passed
    it off to software, who asked me to send them a dump tape, and now
    CSC can't even find the log number for the problem (while it was
    being passed around, I lost track of the log #).  In the meantime
    I have shut down PSDC to prevent further crashes and cannot analyze
    performance very well now that the semester has started and the load
    has gone way up.  Thank you, Digital.

4.  If you use tapes, you will want to know that MOUNT (even BACKUP)
    has been broken so that it waits up to five minutes before issuing
    a mount request to the operators, so they know what tape to mount.
    Not very handy for backups, or user tape mounts either.  "Engineering
    has acknowledged the problem and expects to address it in a future
    release".   With this, I note it was almost 12 months between the last
    two releases for OpenVMS AXP, so we may need to live with this for
    another 12 months.  Thank you, Digital.

    This problem is described in a DSNlink article that is not available
    for viewing by mere customers.

      [OpenVMS] AXP MOUNT Command Takes 5 Min. For OPCOM to Reply to Operator

    I've noticed that DSN ITS only seems to contain solved problems,
    not outstanding ones, which reduces its usefulness.

5.  With this version of VMS, the network drivers attempt to discover
    "misconfigured" networks by using a Duplicate Address Test (DAT).
    It sends out a special packet, and watches for it to come back,
    especially on other devices.  In our case we have an ethernet and
    an FDDI card, and the two networks are bridged by a router.  When
    initializing the FDDI card, even if only for MultiNet, the DAT
    found the packet coming back and disabled the card.  However, if
    we then unplugged the fiber optic cable, counted to 30, and plugged
    it back in, then the card would work just fine.  Digital does not
    consider this a problem.  They claim the router is not handling
    DECNet correctly, but in fact the DAT packet is not in a DECnet
    format.  We finally re-programmed to router to filter out this
    packet, and all's well again.  Thank you, Digital.

    Message to look for: IVADDR, Invalid media address
    (no, it doesn't tell you the address, and it does not get
    logged in the error log)

6.  NPAGEDYN requirements may have increased, but I can't be sure
    because I can't run PSDC.  After analyzing other crashes (about one
    per day initially) I had noticed lots of expansion, and doubled
    NPAGEDYN before shutting down PSDC.  Thank you, Digital.

7.  As you may be aware, Digital feels that tape drives are something
    that only system managers need to use.  Default access has changed
    so that peon users cannot access them, unless you change the
    security profile to allow this.  I believe this was in 6.0 but
    AXP owners didn't see this until 6.1.

8.  On AXPs, and probably VAXes in the future, there is no longer a
    standalone backup.  You can either boot a minimal VMS off of the
    CD-ROM (*very* slow, and don't you want to keep your only copy
    of the CD in a safe place?), or you can install this minimal
    VMS onto a "data disk" (i.e. not the system device).  This
    sort of cramps small systems with one disk; luckily we have
    none.  When you boot this minimal VMS and run BACKUP, if it
    hits a problem such as write-protected tapes, it will use the
    standard OPCOM mechanism to complain and wait for a reply.  But
    if you started the backup at the console, you can't reply, because
    it is busy running the stalled backup, and there is no network to
    allow you to log in elsewhere to do the REPLY.  Thank you, Digital.
    The best workaround is to remember to use /NOASSIST, or else you
    can use control-Y and SPAWN, REPLY, log out, and Continue.
    Of course this minimal VMS also has the problem of waiting five
    minutes before telling you what to mount (a problem if you use
    a procedure that computes which tape to mount based on date).
    (I seem to remember Micro-VMS defined MOUNT as MOUNT/NOASSIST,
    perhaps to "cure" this problem, and that some commentary about
    this still exists today in SYSTARTUP_VMS.TEMPLATE that claims this
    is in SYLOGIN.COM but it isn't any more).

    The good news is that you can finally write a command procedure
    to do your standalone backups, and you can actually get a journal
    file, and run many other DCL commands.

    (Before you get any ideas, the 6.1 upgrade will delete your 1.5
    SYSE directory from the system device).

9.  Don't forget to re-install MultiNet if you have it, preferably
    3.3B, and at least one patch to MultiWare/NFS if you have it.
    It won't otherwise run, as the driver code is different for VMS 6.1.

10. You may want CSC patch 2070 which corrects problems in
    the math run-time library.

11. I had a bit of a problem with an application that ran captive and
    did a SPAWN.  The release notes discuss either adding the TRUSTED
    bit to the SPAWN, or setting Sysgen parameter SECURITY_POLICY to
    71.  It turned out that SECURITY_POLICY is not dynamic, as I had
    assumed, and required a reboot; before I did that, I ended up
    adding the TRUSTED bit to the SPAWN.  My fault; RTFM!

12. Note that if you run SETFILENOMOVE, it won't set the bit on
    files that are open (most of them, because they are INSTALLed),
    generates lots of error messages, and ends telling you:
     ********************************************************
     An error has occurred while attempting to set the NOMOVE
     attribute.  All files may not have been set.
     Please correct the problem and rerun this procedure.
     ********************************************************
    CSC says this is normal, as these files won't be moved if open
    anyway.  (what if I boot minimal, or in some other manner that
    does not cause these files to be installed, and run DFO?).

    In other words, the message "Please correct ... and rerun ..."
    does not mean to correct and rerun.  Thank you, Digital.

There are of course many good-news items, such as being able to run
DFO and BASIC, and we also now run the latest BookReader that can
do searches (thank goodness for that!).  I found the new "Creator"
ACE nice.

It's a shame that such a smooth upgrade left so many jagged edges.
Others beware!

BTW I'm not too unhappy with the support from Digital (except the PSDC
crashes), but was amazed that there were so many severe problems in
the new version of VMS, problems that should have been caught before
public release.  Or did they do major changes after the Beta release?
(naughty, naughty!).

Jim Kirkpatrick
University of Wyoming

