Bob, on A44RTR is see SG1 flapping as well. After breakfast I'll see whether there is a pattern. Initially I thought it was the linux laptop going to sleep after a whule. That may not be the case.
Hans
------Origineel bericht------
Van: Bob Armstrong
Afzender: owner-hecnet at Update.UU.SE
Aan: hecnet at Update.UU.SE
Beantwoorden: hecnet at Update.UU.SE
Onderwerp: RE: [HECnet] netowrk flapping....
Verzonden: 10 juni 2012 01:32
Yup, that's basically a deficiency I've got...
Yeah, but the next question is "Are you the only one", or do other
people's routers have similar timeout issues?
Bob
On 06/08/2012 10:17 AM, Bill Pechter wrote:
My wife refers to the 11/7xx boxes I worked on as my old girlfriends.
;)
If one more piece of hardware
hits the house I'll need the same lawyer.
I'd sure miss Wifey...for a little while. ;)
As soon as I can find it a
good home I'm getting rid of my Vaxstation and going
to go to only emulated hardware under VMware Workstation/ESXi/XenServer at home.
Let me know if you need an extra pair of hands with VMware ESXi; I run
a big host machine here and several more for clients.
And...What model of VAXstation? I'd very happily give another VAX (or
two, three..) a home.
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA
On 06/08/2012 11:08 AM, Gregg Levine wrote:
At one point in time, the OS for the Mac did speak natively to the DEC
family of hardware. It would be very interesting to find out how they
did it. This would greatly benefit Sampsa at least.
Yes, but *which one*? The "original" MacOS bears no resemblance to,
and shares no code with, the current UNIX-based OS. I know of no
MacOS-X-based DECnet implementations. (sadly)
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA
This end is a FVS318...
-Steve
-----Original Message-----
From: owner-hecnet at Update.UU.SE
[mailto:owner-hecnet at Update.UU.SE] On Behalf Of Bob Armstrong
Sent: Saturday, June 09, 2012 18:45
To: hecnet at Update.UU.SE
Subject: RE: [HECnet] netowrk flapping....
and Bob A has a home gw (forgot what it
was) that he claims do the right thing, (not decnet routing..)
Well, I have a Netgear FVS338. It's a "SOHO" box -
somewhere between a
turnkey home router and a fancy Cisco box. I never thought
of it as all
that great, but it does allow me to set up static routes. In
particular I can map specific external ports/Internet IPs to
internal ports/IPs independent of the NAT.
Bob
Any chance this device can be replaced with something IP compatible?
(What's the uplink interface?)
I don't know of a device that I could easily replace it with that is still supported, has VLAN support with multiple virtual interfaces, and without being too prohibitive for a home router.
The uplink port is a gigabit port connected to my FiOS ONT (connected at 100 Mbps).
--Marc
With my connection, I noticed that the circuit would disconnect and
reconnect periodically. It corresponded to the timeout in my firewall
causing the UDP association to be lost. When I increased the timeout
in my firewall for these port 700 UDP "connections", that made my
circuit much more stable.
Like Peter said, Peter and I had a big debate about this at lunch today.
UDP is, by definition (or so I thought), both stateless and connectionless.
I can't understand what state or connection is being timed out in this
case....
Yes, a UDP session is connectionless, however when a firewall is doing
NAT and/or PAT (remember I mentioned that my firewall is randomizing the
source port number, so the LAN port numbers are different from the ones
sent over the internet), it needs to maintain a session table to keep
track of which IP addresses and port numbers map to which systems and
port numbers locally. Those connections time out after a while, and then
subsequent UDP packets wouldn't be recognized.
Any chance this device can be replaced with something IP compatible?
(What's the uplink interface?)
--P
Yup, that's basically a deficiency I've got...
Yeah, but the next question is "Are you the only one", or do other
people's routers have similar timeout issues?
Bob
Ok, I looked thru 36 hours of OPERATOR.LOG on LEGATO for listener receive timeout adjacency down/adjacency up ( flapping as Peter calls it). Here s what I found
The Multinet connection to SG1 flaps all the time, almost like clockwork, with a period of about 20 minutes.
The Multinet connection to GORVAX flapped once in that same 36 hour period.
The Multinet connection to STUPI never flapped, although it did suffer from one corrupted packet error.
The Multinet connections to CIERE and FRUGAL never, ever, flapped nor suffered any corrupted packets.
So, what s it mean?? Bob
I guess that's the issue - you (or your router) really needs a way to
statically define these associations (at least in some specific cases). I
sort of assumed all routers could do that, but maybe I expect too much.
Yup, that's basically a deficiency I've got... well, that and the fact that I can't stop it from changing my source port numbers. I needed this kind of router for VLAN support, so I'm missing a few features that lower-end models would offer like UPnP.
--Marc
it needs to maintain a session table to keep track of which
IP addresses and port numbers map to which systems and port numbers
locally.
I guess that's the issue - you (or your router) really needs a way to
statically define these associations (at least in some specific cases). I
sort of assumed all routers could do that, but maybe I expect too much.
Bob
With my connection, I noticed that the circuit would disconnect and
reconnect periodically. It corresponded to the timeout in my firewall
causing the UDP association to be lost. When I increased the timeout
in my firewall for these port 700 UDP "connections", that made my
circuit much more stable.
Like Peter said, Peter and I had a big debate about this at lunch today.
UDP is, by definition (or so I thought), both stateless and connectionless.
I can't understand what state or connection is being timed out in this
case....
Yes, a UDP session is connectionless, however when a firewall is doing NAT and/or PAT (remember I mentioned that my firewall is randomizing the source port number, so the LAN port numbers are different from the ones sent over the internet), it needs to maintain a session table to keep track of which IP addresses and port numbers map to which systems and port numbers locally. Those connections time out after a while, and then subsequent UDP packets wouldn't be recognized.
--Marc
Marc Chametzky wrote:
With my connection, I noticed that the circuit would disconnect and
reconnect periodically. It corresponded to the timeout in my firewall
causing the UDP association to be lost. When I increased the timeout
in my firewall for these port 700 UDP "connections", that made my
circuit much more stable.
Like Peter said, Peter and I had a big debate about this at lunch today.
UDP is, by definition (or so I thought), both stateless and connectionless.
I can't understand what state or connection is being timed out in this
case....
Bob
The Internet between two end-systems is by arkitekture completely state less when it comes to
knewing anything about what the packets are all about... Unfortenly people love to break that
model to *add value*... -:)
-P
Marc Chametzky wrote:
With my connection, I noticed that the circuit would disconnect and
reconnect periodically. It corresponded to the timeout in my firewall
causing the UDP association to be lost. When I increased the timeout
in my firewall for these port 700 UDP "connections", that made my
circuit much more stable.
Like Peter said, Peter and I had a big debate about this at lunch today.
UDP is, by definition (or so I thought), both stateless and connectionless.
I can't understand what state or connection is being timed out in this
case....
Bob
and Bob A has a home gw (forgot what it
was) that he claims do the right thing, (not decnet routing..)
Well, I have a Netgear FVS338. It's a "SOHO" box - somewhere between a
turnkey home router and a fancy Cisco box. I never thought of it as all
that great, but it does allow me to set up static routes. In particular I
can map specific external ports/Internet IPs to internal ports/IPs
independent of the NAT.
Bob
With my connection, I noticed that the circuit would disconnect and
reconnect periodically. It corresponded to the timeout in my firewall
causing the UDP association to be lost. When I increased the timeout in
my firewall for these port 700 UDP "connections", that made my circuit
much more stable.
Unfortunately, my firewall (a SonicWALL NSA 240) is also stupid in that
it *must* randomize the source port for outgoing packets, so I'm not
able to connect to HECnet because MultiNet insists that the source port
must also be 700 and mine are coming through with random port numbers.
--Marc
IP and UDP is connection-less.. -:)
Throw the firewall away. Find a real router that can do DECnet routing
and NAT and Firewall somwhere. ..
We had this discussion at DCL, and Bob A has a home gw (forgot what it
was) that he claims do the right thing, (not decnet routing..)
-P
Do you know what is the equivalent file for TOPS-10? Is that
SYSJOB.INI?
opr.ato
Understood.
Now I've got another problem. The TOPS-10 node goes yo-yo:
$
%%%%%%%%%%% OPCOM 9-JUN-2012 13:58:58.19 %%%%%%%%%%%
Message from user DECNET on BITXOV
DECnet event 4.18, adjacency down
=46rom node 7.60 (BITXOV), 9-JUN-2012 13:58:58.19
Circuit QNA-0, Adjacent node listener receive timeout
Adjacent node =3D 7.80 (BITXT1)
$
%%%%%%%%%%% OPCOM 9-JUN-2012 13:59:09.10 %%%%%%%%%%%
Message from user DECNET on BITXOV
DECnet event 4.15, adjacency up
=46rom node 7.60 (BITXOV), 9-JUN-2012 13:59:05.04
Circuit QNA-0, Adjacent node =3D 7.80 (BITXT1)
=20
The TOPS-20 one is connected to the same virtual bridge and it is =
running in the same virtual machine, but it runs OK. Any idea? (My =
suspect is the NI configuration in the KHL ini file...)
It's outside tops10, I guess, and I'm completely lost there. If my
braincells starts to work I will rember how to log all the DECnet
packets on the -10 side.
--P
I see these message about every 20 minutes from random nodes. The error
message is almost always:
"Unexpected packet type"
A: Something is corrupting the packet, or they are out of order or a
packet disapeared.
B: Someone is sending packets to you that looks like Multinet Decnet,
but not one of your "peers".
So where do we go from here?
We need more information, and if we can se a pattern.
A: Is there any Multinet link that is stable?
B: Is it related to a site?
If we could do TCPDUMP on this wire with NTP based timestaps on both
sides, we can se what get's corrupted or disappears?
As per subject:
6.3 ITANIC (OpenVMS 8.4 I64)
6.11 BGATES (Windows XP Pro / Pathworks32)
--
Mark Benson
http://DECtec.info
Twitter: @DECtecInfo
HECnet: STAR69::MARK
Online Resource & Mailing List for DEC Enthusiasts.
On 06/09/2012 10:38 AM, Mark Benson wrote:
I had a problem with lacking memory (1GB) on my zx6000
Itanium machines do demand a LOT of RAM. I can only assume the
assembly language/machine code is very inefficient ;)
It's likely data alignment constraints eating up more memory.
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA
On 06/09/2012 03:44 PM, Kari Uusim ki wrote:
Great for heating your house and rapidly running up your electricity
bill, I dare say, mind you I don't expect excess heat is so much of an
issue in Finland! ;)
Quite right - between September and May. We do have hot summers now and
then.
I rather heat my house using computers than radiators. That way I'll get
double the gain. :)
This is good thinking. I recently moved from Florida to western
Pennsylvania. Florida has been getting hotter every year (that climate
change that isn't happening) and the climate here was a significant
factor in our destination decision. The winters here can be quite cold.
Steam boilers are quite efficient at generating heat, but then so are
PDP-8s! 8-)
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA
With my connection, I noticed that the circuit would disconnect and
reconnect periodically. It corresponded to the timeout in my firewall
causing the UDP association to be lost. When I increased the timeout in
my firewall for these port 700 UDP "connections", that made my circuit
much more stable.
Unfortunately, my firewall (a SonicWALL NSA 240) is also stupid in that
it *must* randomize the source port for outgoing packets, so I'm not
able to connect to HECnet because MultiNet insists that the source port
must also be 700 and mine are coming through with random port numbers.
When I worked with them, I found that you could always rely on a firewall
to mess something up somewhere :-(
Right now, I've got no firewall (although I do have NAT in my DSL router)
and I can get everything I want to work correctly with one small exception.
I never got that close when there was a firewall in the way!
I have a Multinet link and I ocassionally see the "Unexpected packet type"
error. However, I've had only 26 of them since November last year. If others
are seeing far more of these errors and are also running firewalls, I would
suggest looking closely at what their firewalls are up to.
Regards.
Peter Coghlan.
On 9.6.2012 22:24, Mark Benson wrote:
On 9 Jun 2012, at 18:46, Kari Uusim ki wrote:
My zx6000 has 2*1,5GHz CPU's, 24GB RAM (just aqcuired), Radeon 7000AGP, but just 4,3GB disks (RZ1CF). When I bought it, it was without RAM and disks.
Great for heating your house and rapidly running up your electricity bill, I dare say, mind you I don't expect excess heat is so much of an issue in Finland! ;)
Quite right - between September and May. We do have hot summers now and then.
I rather heat my house using computers than radiators. That way I'll get double the gain. :)
VMS is installed on those small disks and that wasn't difficult. I didn't need a shoehorn :P . VMS really don't need enormous disks - even if we are talking about the Itanium version.
But to be honest, I have to find bigger and faster disks. :)
I hit lucky just after I got mine and found a guy selling new, boxed and original HP 15krpm 73GB drives in the correct carriers for the zx6000 for 60 each shipped. I grabbed 2, one is in my IBM POWER5 machine (because it matches the existing disk) and the other is in my zx6000.
On 9 Jun 2012, at 18:46, Kari Uusim ki wrote:
My zx6000 has 2*1,5GHz CPU's, 24GB RAM (just aqcuired), Radeon 7000AGP, but just 4,3GB disks (RZ1CF). When I bought it, it was without RAM and disks.
Great for heating your house and rapidly running up your electricity bill, I dare say, mind you I don't expect excess heat is so much of an issue in Finland! ;)
VMS is installed on those small disks and that wasn't difficult. I didn't need a shoehorn :P . VMS really don't need enormous disks - even if we are talking about the Itanium version.
But to be honest, I have to find bigger and faster disks. :)
I hit lucky just after I got mine and found a guy selling new, boxed and original HP 15krpm 73GB drives in the correct carriers for the zx6000 for 60 each shipped. I grabbed 2, one is in my IBM POWER5 machine (because it matches the existing disk) and the other is in my zx6000.
--
Mark Benson
http://DECtec.info
Twitter: @DECtecInfo
HECnet: STAR69::MARK
Online Resource & Mailing List for DEC Enthusiasts.
My zx6000 has 2*1,5GHz CPU's, 24GB RAM (just aqcuired), Radeon 7000AGP, but just 4,3GB disks (RZ1CF). When I bought it, it was without RAM and disks.
VMS is installed on those small disks and that wasn't difficult. I didn't need a shoehorn :P . VMS really don't need enormous disks - even if we are talking about the Itanium version.
But to be honest, I have to find bigger and faster disks. :)
Kari
On 9.6.2012 20:06, Mark Benson wrote:
On 9 Jun 2012, at 17:21, Rob Jarratt wrote:
That *is* cheap, but wrong continent :-( I will look around for these now.
Regards
Rob
There's one rx2600 on Ebay right now, check item #200773544925.
That's as cheap as it gets.
Damn, I need one of those server front-panels ;)
Almost all Itanium machines can run VMS.
See http://h71000.www7.hp.com/openvms/hw_supportchart.html
I didn't realise it was that widespread, although most of those are later model Integrity servers that came *after* the rx2600, and so are likely outside the Hobbyist ballpark. Maybe I was confusing the framebuffer support or something.
If you are feeling RICH there's always eBay #170855134495 - Nice lump of hardware for that money but you;d need a SAS 2.5" disk which would cost nearly as much as the server :P
The workstation models are unfortunately rare.
... and when you find zx6000s they usually have FireGL4 cards in like the one I just ripped out of mine. No VMS support for those and HP-UX support is limited only to the specific distribution supplied with the zx2000/6000 workstations :(
Radeon 7000 and 7500 cards for Itanium are rarer than the machines!
Sometimes you might find one on Ebay. The advantage of the workstation models are that the fans are more quiet than the server models have.
You can tell if a machine has the zx6000 fans, visually, because they are only half depth compared to the rx2600 server fans (so only fill the front half of the hot-swap fan bay). As Kari says, the zx6000 fans are much quieter, probably not so droning as server fans either. My zx6000 is quite amicable to sit next to with 1x 15krpm 73GB disk, however I yanked a CPU out ogf mine so it only has one CPU turbo-fan running which cuts the noise down quite a bit.
Just some food for thought... My zx6000 came orgiinally configured with 24GB of RAM and 2x 1.3GHz CPUs and a FireGL4 graphics card. The original purpose? It was a CAD station for Honda Formula 1 at Bracknell :) I think Mark Wickens has one (but his has all the outer skins still, mine doesn't) fropm the same place and so does Dan Williams. for machines that date back to 2003... that's a lot of spec!