http://h71000.www7.hp.com/?

Bob Gezelter · Posted: Mon Jul 17, 2006 2:54 pm Post subject: Re: http://h71000.www7.hp.com/?

To all,

I have heard back (and verified) that the site is once again reachable.

Apparently, the problem was a network infrastructure involving a
switch. The OpenVMS server was sitting there, unreachable.

- Bob Gezelter, http://www.rlgsc.com

Dave Froble · *nix forums Guru Joined: 21 Jul 2005 Posts: 1172

Bob Gezelter wrote:

Mark Daniel · *nix forums beginner Joined: 01 Mar 2005 Posts: 44

Bob Koehler wrote:

Thomas Wirt · *nix forums beginner Joined: 29 Jul 2005 Posts: 25

Dave Froble wrote:

Bob Koehler · *nix forums Guru Joined: 25 Jul 2005 Posts: 1078

In article <12bl6fh34ioto53@corp.supernews.com>, Mark Daniel <mark.daniel@vsm.com.au> writes:

Thomas Wirt · *nix forums beginner Joined: 29 Jul 2005 Posts: 25

Thomas Wirt wrote:

Dave Froble · *nix forums Guru Joined: 21 Jul 2005 Posts: 1172

Mark Daniel wrote:

Hoff Hoffman · *nix forums Guru Joined: 22 Jul 2005 Posts: 572

Bob Koehler wrote:

warren sander · *nix forums beginner Joined: 11 Feb 2005 Posts: 24

Trying to handle everything

1) ping et al will not work for any hp.com type node cause the firewall's
block all that stuff. And you can only
traceroute to the firewall.

2) the server is a DS20E 833 dual processor, 2GB memory running OpenVMS 7.3
and Apache 1.3.
It's running 7.3 because of access issues with updating the OS on 3 systems
at the same time (my development, mirror and deployment systems). If I
upgrade I need to update a systems in 3 seperate facilities to the same OS
version before I mirror any content out on the off chance that a script etc
stops working (And a couple of my scripts think they know stuff about the
inside of VMS that they really shouldn't but were written before some of the
current interfaces were created and since they work they haven't been
re-written). I've been waiting for approval of Integrity systems to replace
the AlphaServer's so I can start anew but that's taking more time than I
have wanted (3 years so far).

3) The problem is the switch that the externals server is connected to. It
went out last month on a weekend also and didn't sound the alarm. They just
rebooted it last time as it was the first time. This time some more looking
around and scratching of heads leads them to not knowing why it didn't cause
an alarm which would have gotten attention and the off-air would have only
been about 20 minutes on a saturday morning. They are going to replace the
switch tomorrow AM (probably just move the cables someplace else) and figure
out off-line the alarm issue.

4) I think I have one of only a couple of OpenVMS systems connected to HP's
external network (the testdrive systems are the only other ones I know
about).

5) the original OpenVMS server was based on OSU in 1995 and was an illegal
cluster configuration of 2 AS200 4/233 (mustang) systems SCSI clustered to a
BA232 pizza box. One of the mustangs was donated to the Digital Internet
group as the original digital.com search engine running altavista software.
Not the altavista main server but just an search of the digital.com sites
(like search.hp.com is today). Anyway we broke up the cluster cause we never
failed over in 2+ years decided that one system was ok and a site search was
a better use of the resource. Then we got AS1000's to replace the mustang
and sometime in there we switched over to purveyor as the web server. When
they got CSWS done I was asked to switch over to apache which I did.
About 4 years ago the Tru64 UNIX AS1000 started having some memory issues.
And there were a number of DS20's scheduled to be crushed. Working with the
folks that host the servers we managed to save a couple of the DS20's and
use them for the Tru64 UNIX and OpenVMS servers. Which is where we are
today.

6) JF the only logical conclusion that you can make is reliance on
technology to alert you to problems and not providing a method to 'ring the
alert yourself' is an issue that HP and many other companies are finding
themselves having with reduced staffing levels and higher needs for service
levels. This has been taken care of at this point.

7) the server has been up for 13 days (I had to restart it to add another
server to the system (Yes I run more that just he OpenVMS server on this
hardware). The Alpha has been up for 132 days I don't remember why it was
rebooted. It was either patches or a power shutdown in the facility or a
power failure. Before that it was up over a year.

8) back in 1995 we put together some pages that would let you execute a
couple of show commands:
http://h71000.www7.hp.com/cgi-bin/show-sys.exe/0 does a show cpu/full
http://h71000.www7.hp.com/cgi-bin/show-sys.exe/1 does a show system
http://h71000.www7.hp.com/cgi-bin/show-sys.exe/2 does a show memory
http://h71000.www7.hp.com/cgi-bin/show-sys.exe/3 does a show cluster
or go to http://h71000.www7.hp.com/misc/about-server.html

"Hoff Hoffman" <hoff-remove-this@hp.com> wrote in message
news:51Sug.624$Ir3.562@news.cpqcorp.net...

Mark Daniel · *nix forums beginner Joined: 01 Mar 2005 Posts: 44

warren sander wrote:

w_tom · *nix forums beginner Joined: 19 Mar 2005 Posts: 28

Two principles to first understand how switches are damaged. First,
all switches already contain effective protection. Second, protection
may be overwhelmed if a transient finds a path to earth via that
switch. Nothing - not a protector, not a fuse, not a magic box - will
stop block or absorb the destructive surge. And yet we routinely
suffer direct lightning strikes without damage - see the telco example
provided later.

Solution is also found in a principle pioneered by Ben Franklin in
1752. Don't even try to stop or absorb a surge. Shunt (divert) it.
That is what a shunt mode protector does. Lightning seeks earth
ground. If you don't earth a transient where it enters the building,
then transient will find destructive paths via electronics.

Destructive surges are electricity. Electricity means both an
incoming and outgoing path must exist. If a surge enters on a switch
and has no outgoing path, then no electricity flow; switch is not
harmed. However all switches have numerous incoming and outgoing
paths. Incoming on AC electric (the most common incoming path) and
outgoing via signal lines - especially those line connected to telco
wires.

Telco lines already have an effective 'whole house' type protector
connected short to earth where their wires meet yours. Therefore phone
lines are typically not an incoming path. But that telco 'provided for
free' 'whole house' protector makes a good outgoing path. Incoming on
AC electric; outgoing on phone line. Just one example of how
electronics are routinely damaged. Parts that often fail are on DAA
(telephone wire) side of that path.

Solution is not found in adjacent protectors. Solution is to earth
every incoming utility wire to a common (single point) earthing
electrode AND make that earthing connection as short as possible (ie
'less than 10 feet'). That means all utilities must enter at a common
location - to be earthed short to the same earthing electrode.
Distance is critical which is why wall receptacles are not sufficient
as earthing.

'Whole house' protectors for AC electric are sold in Home Depot,
Lowes, and electrical supply houses. These effective solutions are not
sold in Radio Shack, OfficeMax, Kmart, Sears, Staples, Circuit City or
other sources of ineffective protectors. Effective solutions have
responsible names such as Siemens, Square D, Cutler-Hammer, Leviton,
Intermatic and GE. Effective protector has a wire for the dedicated
earth ground.

Even some RJ-45 ethernet protectors have that all so necessary
earthing wire. But again, a protector is only as effective as that
connection to and quality of earth ground. Earthing - not the
protector - determines whether appliance internal protection is
overwhelmed. You had damage from a nearby strike (which is really more
often a direct strike)? Then look at your protection system.

Where is the path to earth? Through a non-destructive path (as Ben
Franklin demonstrated) or via electronics? Most critical component of
every protection 'system' is the component that everything else centers
upon - a single point earth electrode.

Above defines secondary protection. Primary protection 'system' also
must be inspected:
http://www.tvtower.com/fpl.html

Why does damage occur? Because a transient was not earthed where
wire enters a building. Therefore the transient found a destructive
electric circuit - incoming and outgoing path - through electronics.
Your telco connects its $multi-million computer to overhead wires
everywhere in town. Do they shutdown for a thunderstorm? Of course
not. Is that computer damaged by thunderstorms? It must not fail.
What do they do? Same thing. A good earth ground AND all incoming
wires first connect to that earth ground via a 'whole house' type
protector and before entering the building. A technique well proven
for maybe 100 years. A technique so well proven that surge damage is
considered human failure. It is that routine to eliminate. Switches
failing? Start by inspecting the earth grounds for each - the primary
and secondary protection 'systems'.

One poster in the 15xxx zip code may live where geology tends to make
transient damage from every ground strike most destructive. Damage is
often defined by geology. Some therefore must make the single point
building ground so large as to enclose the building inside either a
halo or Ufer ground. Again, protection is defined first and foremost
by the quality of earthing. Damage being a human failure. No earth
ground means no effective protection. No way around technology that
was standard even before WWII.

Thomas Wirt wrote:

Paul Sture · *nix forums Guru Joined: 12 Feb 2005 Posts: 357

Peter 'EPLAN' LANGSTOEGER wrote:

JF Mezei · *nix forums Guru Joined: 21 Jul 2005 Posts: 2556

Dave Froble wrote:

Dave Froble · *nix forums Guru Joined: 21 Jul 2005 Posts: 1172

w_tom wrote:

w_tom · *nix forums beginner Joined: 19 Mar 2005 Posts: 28

Little difference whether a cable is overhead or underground. An
example. Lightning strikes the other building. Lightning seeks earth
that is some four miles distant from that building. Your building is
in that path. What is the shortest path, electrically, from that
building to earthborne charges four miles away? Via that buried cable
and through electronics inside your building, then to earth, and onward
for four miles. Your electronics are now in a destructive path that
included buried cable.

EMP is so trivial that even an NE-2 neon glow lamp (a bulb rated in
milliamps) will make such transients irrelevant. If a transient
overwhelmed protection already inside appliances, this it was not EMP.
It was a direct strike.

Every incoming wire in every cable that enters a building must
connect to a common earthing electrode, either by direct (hardwire
connection) or via a protector. In most buildings, only one AC
electric wire is grounded. Other AC electric wires are ungrounded (due
to no 'whole house' protector) and therefore connect nearby lightning
directly into electronics. With only one AC electric wire grounded
means AC electric is not sufficiently earthed. Most all buildings do
not have sufficent earthing ('whole house' protectors) on AC electric.
No 'whole house' protector on AC electric means AC electric wires are
not properly earthed.

This figure from an industry professional demonstrates two structures
with single point ground. Even a buried wire must connect to that
earthing before entering a building.
http://www.erico.com/public/library/fep/technotes/tncr002.pdf

Electronics already contain internal protection. Protection that can
be overwhelmed if a destructive surge is not earthed before entering
the building. My apathy for adjacent protectors is because anything at
the electronics that will protect those electronics is already inside
electronics. Experience has demonstrated how adjacent (and not
earthed) protectors even contributed to damage of adjacent (and powered
off) electronics.

Many reasons why telco lines are not 'most often' struck by
lightning. One is that AC electric lines are above telephone lines and
therefore protect those phone lines. A most common source of damage to
telephone appliances are transients that enter on AC electric. There
are exceptions. But this typical example also demonstrates how to
identify those exceptions. Bottom line - the analysis must answer how
lightning traveled to get to earth. The solution must change that path
so that the transient finds earth without entering the building.
Earthing is THE most critical protection component in a building wide
solution.

Dave Froble wrote:

Google