|
This text is largely based on Peter Wendt's posts.
1st Rule you learn: Station ports are for stations only, Ring In / Ring
Out
for RI/RO only. Especially when it comes to the "dumb" (but rock-solid)
8228.
>They had me use the setup aid in the bad port
with the network live at
that time, hence why I thought that doing this
was OK. However using the setup aid did not fix the port and we had
to move the plug to a free port ( the last one free) on one of the other
MAUs.
Aha. The self-acclaimed "experts" weren't any better. It's some sort
of a miracle that the entire net didn't go down. Usually "breaking the
ring" with a setup tool causes the other ports to shut down and drop the
relays.
In either case you cannot test a bad port / relay whatever
with the network on. Simply not possible due to the "double-ring" structure
of TR. What morons !
>I pulled the plug labeled receiving in the LAN
closet and inserted the setup aid. Once the setup aid finished (red
light blinked) the 2 lanmanager servers that run the POS system (plugged
in on the POS closet) started beeping and reporting beaconing.
Awww. Great. You *may* try to disconnect a station for about one - two
seconds e.g. while changing to a different port by what reason - or
to a
different MAU on the same ring - but once you hold it off the ring
for more
than 3 or 4 seconds it will surely collapse. In most cases the TR interface
does not open again. It will maintain the pulse (phanto voltage) for
2
seconds probably, which forces the "new" port open - but after that
it is
ordered to shut the port from the card hardware. Uh-oh.
>I was told that the 2 closets were on separate
loops by the staff at work
that was supposed to know how to maintain the
POS system. It must not be, otherwise the POS system would not have
went down with the LAN loop. All could have been avoided if someone
had actually documented something correctly, all I had to work with was
a 4 year old diagram of the POS loop which must have changed since they
put in the second closet ( and never bothered to document the changes ).
Sigh. What more to say ? It's the usual stuff. Everywhere - all the
time.
>Is it possible to connect loops through one of
the 1-8 ports on a 8228, or would it have to run through either the RI/RO
ports?
No. You cannot. Not that easy. There were "station splitters" available
from e.g. Olicom that allowed to abuse a "station port" for another "stub
8228", which holds "another subset" of stations.
Basically and directly you cannot a) connect 2 station ports of
two
different 8228s to make "one net from formerly two nets" nor b) connect
a RI or RO port from an 8228 to a data station port. It lacks the phantom
voltage for the relays on these ports. Only 4 data wires are present there,
which are 2 by 2 shortened by default and automatic (ICS connectors).
For a) you need to RI/RO daisy-chain them. Usually no problem - even
in
online mode. For b) you need an intelligent MAU like the 8238 where
port # 16 can be configured to act as RI/RO chain for 8228s (but the stations
attached on the 8228s show up as attached to port #16 in "the inventory
list" on the 8238).
>Do you think that the troubleshooting I went
through the first time could
have changed, caused or contributed to the problems
that I had noticed
recently ( workstations randomly dropping from
the net )? I think there was, if not still is some problems
with the setup that is currently running as the MAUs in the LAN closet
tend to cick-clack too much for workstations that are always on and not
turned off ( also something the people who maintain the POS system say
is normal when I asked about it "oh they always make that noise allot"
).
Ugh ... not true normally. You *sure* hear the relays click on / off
when a
station enters / leaves the ring - but for some period of time there
should
be "deafening silence" in a MAU room full of 8228s with no additional
components, routers, switches etc.
>Lastly to note is that once I had unplugged all
the nodes from the LAN
closet ( I left the RI/RO ports in on all the
MAUs though) and reset each port all the workstations came back up including
the one in receiving that prompted the whole thing.
That's the consequent method to get it done. Usually all alone in the
building, during the night and with some spare 9V battery blocks at
hand ...
Been there - did that. Drove to a 24h filling station at 3 in the morning
after my 3 or 4 MAU tools exhausted their batteries.
>I didn't test the dead port that I had found
the first time though, it makes me wonder if I happened to fix that as
well this time around.
Well - "Why not ?" You do them all - or none. If there is one really
defective port there you have the potential danger of running into the
same scenario once again later. There are *rare* incidents where an 8228
is really defective - but I *had* that. And you can only find out with
testing *all* ports ... best with the RI/RO cables detached too and a second
Tester in the RI or RO port. When the lamp on the station port stays out
the RI (or RO) port *must* come on. That's the indication that the 8228-internal
data path is working. You can rely however on the "click-...-clack" noise
fo the station relay only and hope everything internally is well-fabricated.
Most are - working fine for decades. Some are not. We had a number of Brazil-made
8228s, which had an internal fault and did not pass the signal to either
RI or RO (or both on one). All station ports clicked fine ....
Causes For Permanent "relay static" in a group
of MAUs
1. Worn out plugs, improperly seated, partially seated
and lacking the
retainer. Largest group.
2. Worn out cables. People walk over them, run over
them with office chairs or place furniture on them. Second largest group.
3. Cable lenght exceeded between two MAU (over 400
/ 200 meters or 1200 / 600 feet at 16 / 4 Mbit/s or longer than 200 / 100
meters or 600 / 300 feet between MAU and station. Improper (to tight) bended
with less than 5x the cable diameter. Happens. Fairly often. Often detected
after components had been replaced or an existing net is switched from
4 to 16Mbit.
4. Defective adapters. The capacitor building up
/ holding the relay phantom voltage is dryed out. Defective connectors,
improperly affixed cables at the adapter end (see 1.).
5. Devices usually *not* in the net with adapters
installed and the LAN
software still active try to insert in ring .... either with false
speed or
get disabled by LAN program timeouts periodically. No joke - happens
at a customer. They had left-over machines with a 4Mbit adapter and had
them yet attached to the MAU. They have been used to copy diskettes, virus-check
files and suchlike. They come on, run 5 minutes - are turned down. Network
interface opens several times ... closes ... causes beaconing ... goes
away ...etc. Took me two days to locate the station with a LanAlyzer and
fetch an address - which was then pointed to an "old machine no longer
in use".
6. Defective MAU. Happens, but is unlikely. Once
you'd "prime set" an 8228 it is good for a few more years. The step should
be repeated after 3 - 4 years in use.
So that's *my* experience of over 13 years with 8228 units ... :-)
Peter in Germany
What Are the Differences between Bridges,
Routers, and Gateways?
A bridge is a device that interconnects two subnets using
the MAC address of stations. Bridges operate at layer 2 of the OSI protocol
stack. Because MAC addresses are protocol independent, a bridge is very
useful for physical network extension, speed matching, and wiring optimization.
However, since a bridge is essentially transparent to the attaching stations,
dependencies upon such protocol variables as timers, flow control and error
recovery are not supported. Thus, problems may arise if too many workstations
are bridged. These include issues such as broadcast storms, time-outs and
congestion-caused delays in response times.
A router is a device that interconnects two or more network
segments on the basis of OSI layer 3 addresses. Because such addresses
have protocol
dependencies, a router must either participate in the protocols for
purposes of routing, flow control, and error recovery, or must encapsulate
protocol dependent information in frames supported by the routers. Even
with encapsulation however, a router must understand the network topology
and where to send data frames. Routers maintain topology information as
a result of an exchange of topology information with other routers in the
network. Communication of topology information between the routers may
use open standards for exchange such as OSPF or EGP or vendor-specific
protocols. They do not assume responsibility for the integrity of the address
information used within the network.
A gateway is a device that handles interconnection and
transfer of information between entities in layers 4 through 7 of the OSI
Reference Model. Thus a gateway is required when an entity using one protocol
needs to communicate with an entity using another protocol (protocol conversion),
when the addressing structure of two entities is inconsistent, or when
connection is required between independent networks.
Direct Cable Connection?
Unlike connecting two BNC net cards, you can't connect
two TR cards together without a hub (apocryphal tales of a magic cable
and software are unfortunately just that AT THIS TIME). The main
problem is two-fold: 1.) the adapter detects the opening of the ring-insertion
relay and 2.) on the "open adapter command" the communication (insertation)
into the ring must be finished within a certain period of time.
A direct cable connection does not supply the required
change of current flow in the DC-phantom voltage line and to manage simultaneously
starting the ring communication on two different PCs *within* the time-limit
is more than unlikely.
Tales of Direct Connection
From Terry King:
Peter, back when I was at IBM in Vermont there was some
internal software you could use to make two token-ring cards talk to each
other without a MAU. Somebody must have done this outside IBM???
Peter's Response:
As I wrote earlier in this relation (but not in this thread): there
were rumors about such a tool - and corresponding to a team-mate "We had
that back in 1990 or so ... required a particular cable with 2 resistors
..." but disk and cable have (of course !) gone.
Funny fact: we once had this software in the company (back
in 1990 or so). Someone brought it there and we had a "magic cable" with
a little black box in the middle, which might have contained some resistors
and caps - but no relay.
On demand I got told that it works this way:
- connect the two computer with that "magical" cable
- start computer #1 ... will come up and "start the ring"
- start computer #2 ... will complain about not being able to insert
in the ring
- start the "magical" software, which will *force* the TR-adapter to
open and insert in the ring
The normal opening requires to receive the "relay opened"
sense signal (change of current flow in the two wires for the +5V phantom
voltage when the relay opens), then trying to read the passing token (if
ring-speed matches ! - or Autosense the ring speed before), then open the
adapter and copy the token informations and pass the token to the next
station (or copy the data from it ..). This procedure must have finished
within a "time-out" limit. If the opening / inserting fails an error-code
is displayed.
The problem is to detect the "ring opened" flow-change
with the MAU missing ... while this cannot be detected the follow-up inserting
will not be started. The software only opens the adapter without waiting
for the "ring opened" signal.
I'm still trying to get more informations about that.
In case I find something ... I will tell.
|