[Info-vax] 8.4 freespace-drift problem?

[Stephen Hoffman]

On 2015-05-22 12:26:57 +0000, Hans Vlems said:

> Hoff, us hobbyists have to live and work (play?) with what we can afford.

The commercial sites have a budget, too.

When I get mail on this stuff, I ask the folks whether they have enough 
of a budget for the hobby.   Can the folks afford to have boards fail, 
and to repair or replace them?   If not, the folks might want to try a 
different hobby, or a different approach.  Without a budget for parts 
and pieces, the hobbyist gets to spend time repairing and soldering 
boards (and with SMT, that's a hassle),  or a whole lot of frustration, 
or sometimes both.   As for a different approach toward hobbyist folks 
that want to use OpenVMS, hardware emulation on an x86 box is the 
cheapest way to run OpenVMS these days, and you can emulate some pretty 
decent configurations with inexpensive x86 parts underneath.

> Given the quirks of an rx2600 I'd rather much prefer a DS20/25.

I've been running Itanium since before it was released.  It's got 
quirks certainly, but so does old Alpha.

For now and for hobbyist folks that want real hardware, I'd still 
recommend scrounging a zx2000 (I've seen those sell for US$200 to 
US$450) if you can find one, but hopefully VSI gets that x86-64 port 
going and we then get a better and more appropriate selection of 
hardware available.   zx6000 and rx2600 can be quiet, but the 
higher-end Itanium boxes are loud.  The decent-grade Alpha boxes are 
getting more expensive lately, and they're also usually much older.

Given the lack of patch access for hobbyists, staying current gets more 
interesting, as you can at least get a recent UPDATE patch 
AFAICT/AFAIK.   For this case, there have been a number of patches for 
V7.3-2, and a number of them are mandatory patches and fixes for 
crashes and related.  These patches for older releases are generally 
not available to hobbyists, too.

> My point is that old supported hardware and new openVMS versions is no 
> problem at all. There have been questions about the quality of 8.4. OK, 
> nuisances happen but altogether this latest version works well.

Early V8.4 was dicy, yes.  Current works well.  OpenVMS I64 UPDATE V5 
or later is locally preferred.

> Even on an EV56 class machine like an AS1200. And even on its white box 
> cousin, or should I say half brother...
> 
> I feel Phillip got bitten by taking a large step, upgrading from 7.3-2 
> towards 8.4 in what appears one move.

Donno.  Maybe.   Clean installs and standard file locations and typical 
configurations are less hassle.  That approach is often viewed as an 
anathema around here, but it works.  After tracing the startup through 
with the current environment, one of the next troubleshooting steps 
would be installing OpenVMS V8.4 on a scratch disk.  Whether that 
works.  Not clustered.  Not tailored.  Not HBVS across 10 MbE or 100 
MbE.  Not tuned or tailored beyond what's in the V8.4 documentation.

Then either hauling over the old files and tools across, or maybe 
falling back to the customized disk and doing some debugging.

I'd then start bringing minimally-tailored V8.4 systems into the 
cluster.  Emulations, or maybe real hardware.   Probably as 
secondaries; as non-voting boxes that depend on the presence of a 
primary voting host.   Since HBVS across 10 MbE or 100 MbE can be very 
slow and most folks want some redundancy around the "common" disk, I'd 
scrounge a supported RAID controller for the core voting host, and then 
just serve that.  This if there's no option for multi-host 
interconnections here.

Trying to deal with Alpha and particularly with the EV5 and earlier 
boxes and with the old SBB 9 GB SCSI disks is going to be more work, 
more time, more effort and more risk.  At its core, this is why many 
commercial businesses replace and roll out the old gear on a schedule, 
and why businesses upgrade to newer software.

Put another way, I'd fully expect to have troubles, hassles, failures 
and weirdnesses with the configuration that Phillip is running.

> And that he ought to run the XP1000 for at least two days and then run 
> autogen with feedback.

I'd clean-install it, too.   If this was an actual-hardware 
configuration and a running-the-house sort of hobbyist environment, I'd 
move to the XP1000 as the core system — or whatever the newest and most 
capable box is within Phillip's hardware menagerie — and use emulation 
for most or all of the rest when and if those other systems are needed, 
and save the rest of the hardware for standalone use on fun days or 
museum days.   I'd probably also move the local DNS services and other 
core services to small x86 or ARM box; a plug computer, micro server or 
something of that ilk.   But this is entirely Phillip's decision.


-- 
Pure Personal Opinion | HoffmanLabs LLC