This article is from the Token Ring FAQ, by James Messer James@NetworkUptime.com with numerous contributions by others.
Phase 0: Media Lobe Check Troubleshooting
Failure to complete phase 0 is one of the most common failures when
trying to configure a token ring network interface card into a PC.
Most token ring adapters, upon failing, will display some cryptic
error message like "Adapter failed to open." or "Failed
initialization.". Always check the cable connected to the adapter
and where it connects to the hub. In order for an adapter to pass
phase 0, it must have a closed circuit to test. Either use a wrap
plug or insure that the adapter is connected to a working MAU. Bad
cabling causes many adapter problems during the insertion process.
Things to look for include: "Is the adapter configured to use the
correct media port, UTP or STP?", "Is the cable run from the adapter
to the hub complete and correct?", "What exactly is between the
adapter and the hub, how many punch downs, what kind of cable, how
is it wired, where does it run, are there phones in the same cable,
etc.?", and "What kind of media filter are you using?". Keep in mind
that what will work at 4 Mbps will not always work at 16 Mbps.
Phase 1: Physical Insertion Troubleshooting
Many of the problems associated with phase 1 of insertion are the
same ones accounted for in phase 0, especially bad cabling and bad
media filters. The error messages at this stage are usually the same
as those received during phase 0 and are just as cryptic. If the
cabling checks out, look at the hub. Does the hub indicate
insertion? Does the hub make a chattering noise when the adapter is
trying to insert? Are there other stations on the ring? The problem
could be cabling or a faulty adapter (not supplying consistent
phantom can cause the relay to chatter). Some simple steps would be
to move the station to a working location or try a known working
station at this location. Can the station in question insert if the
other stations are turned off? It could be that there is a physical
layer problem (i.e. wiring, line noise, jitter, etc.) on the ring
which shows up as more stations insert, causing purges and beaconing
which will kick off a new inserting adapter. If you are sure that
the cabling is acceptable, you will probably need a protocol
analysis trace before making any prognosis as to why you can not
insert. The analyzer should be the immediate upstream neighbor to
the station trying to insert.
A normal insertion that completes successfully commonly causes
several token ring errors on the ring during phase 1. Common errors
at this time would include burst errors, line errors, token errors,
ring purges, and lost frame errors, due to the simple act of opening
the relay. Do not assume that the existence of these errors indicate
a problematic ring, as these are normal symptoms that occur during
the insertion process.
Phase 2: Address Verification Troubleshooting
The only time you need to worry about this phase is when you are in
an environment where the user is using LAAs. When users start
entering LAAs, the chance of duplicate addresses goes up
dramatically. The most common cause is copying a working adapter
configuration files (config.sys, autoexec.bat, net.cfg,
protocol.ini.) between stations. The symptom to look for is when the
adapter is trying to insert, it will (under most circumstances)
insert and de-insert twice in rapid succession and then quit trying.
It will also provide messages such as "Adapter failed to
initialize." or it might actually say "Failed Duplicate Address
Test.". Change the LAA or move to another ring and try to reinsert.
If you can get a trace of the failure to insert, you can look for
the duplicate address test frames. As in phase 1, insert your
analyzer directly upstream to the failing adapter.
Phase 3: Participation in Ring Poll Troubleshooting
Some probing is usually required to find out the root of the problem
at this phase. If you can not insert, time how long it takes for an
inserting adapter to fail. If the answer is 15-20 seconds, then it
is probably failing the ring poll. If the answer is less than 15
seconds, the problem could still be the ring poll failure but more
information will be required.
If you get a trace of a ring that is failing the ring poll process,
you will find a MAC frame issued by the AM called Neighbor
Notification Incomplete (NNI) or Ring Poll Failure. This frame
should be issued every 7 seconds in a failing ring just prior to an
AMP MAC frame. The NNI frame is important because it will contain
the address of the last station to successfully complete the ring
poll process. The downstream neighbor from this station is usually
the culprit and removing the downstream neighbor should cure the
problem. Exceptions will occur if there is more than one station
that is not participating in the ring poll process. Another way to
cure the problem is to have all stations on the ring power down for
30 seconds (at the same time) and then try to reinsert, however,
this is only a temporary cure and not a fix since the problem will
likely reappear. If the failure is proven to be a ring poll failure
and the problem persists, the customer may need to look at
contacting the vender of the failing adapter(s) or device(s) and see
if the vender has a newer driver available.
Phase 4: Request Initialization Troubleshooting
Failure at this stage is rare but could point to a failing adapter
on either the RPS or on the inserting station, a physical layer
problem on the ring (cabling, jitter, etc.), or some other
`undocumented' feature of the environment in which the failure
occurs. The only method to determine a failure at this stage is to
use an analyzer inserted as the upstream neighbor to the adapter in
question. An RPS is generally best serviced by bridges or routers
since they are usually running the server software required to
perform these services.