| Title: | DIGITAL UNIX (FORMERLY KNOWN AS DEC OSF/1) |
| Notice: | Welcome to the Digital UNIX Conference |
| Moderator: | SMURF::DENHAM |
| Created: | Thu Mar 16 1995 |
| Last Modified: | Fri Jun 06 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 10068 |
| Total number of notes: | 35879 |
I have a customer running V4.0A on a 4100 with a PCI Token Ring Adapter PBXNP-AA. Their kern.log file is being flooded with messages of the following type: test4100: vmunix: tra0: ring status: ring recovery (Their machine in named test4100). Once in a while they are having connections to the machine dropping. Other notes in this Notesfile seem to suggest that their are some token-ring patches available, but I have not found them in the public areas on oskits or guru. Customer is willing to upgrade to V4.0B if this will solve the problem or a V4.0B patch exists. (The V4.0B patches seem to be more up-to-date than V4.0A). Thanks, Ram
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 8760.1 | Get the patch | SMURF::GILLUM | Kirt Gillum | Mon Feb 10 1997 13:20 | 10 |
There is a patch for running on the 4100 (a TI errata implementation).
Contact Bob Spear (spear@zk3.dec.com) for a pointer to the patch.
Typically "ring recover" occurs everytime that a node enters/exits the
ring. Also, if you look at the counters, you'll probably see alot of
beaconing (indicative of a bad connection).
However, start with the latest patched driver.
| |||||
| 8760.2 | still errors after patch | DYOSW5::WILDER | Does virtual reality get swapped? | Wed Mar 05 1997 10:09 | 17 |
I am also working with the customer listed in the base note.
The customer is now at 4.0b. We have applied the patch provided by
spear. They are STILL getting a lot of "ring recovery" errors. After
all this, the system crashed after running for about 9 days. The
crash-data showed numerous ring recovery errors at the time of the
crash.
To our knowledge, there are not nodes entering/exiting antwhere NEAR
the number of ring recoveries they get.
Any ideas? Should we ignore this?
Thanks,
/jim
| |||||
| 8760.3 | SMURF::GILLUM | Kirt Gillum | Thu Mar 06 1997 18:06 | 8 | |
Ring recoveries are not a big deal... Perhaps you should try a
different cable/mau port on the adapter. Also, look at the counters
and see if anything abnormal jumps out at you (netstat -s -Itra0).
Crashing is a big deal. Why are you crashing? Is the token ring
driver on the stack when the system crashes?
| |||||
| 8760.4 | Recovery errors concern them | NETRIX::"nancy@csc.cxo.dec.com" | Nancy Flavell | Fri Mar 21 1997 15:05 | 34 |
Kirt, We appreciate your help with this problem. Since installing the latest Digital UNIX patches (including the one you mentioned from Bob Spear), the system has not crashed again. It seems that there is not even a core dump or crash-data left to analyze on the system, which has been running for eight days since applying the patches. However, they still get so many of the "ring recovery" messages that they are quite concerned the Digital equipment or software has some kind of problem. Their viewpoint is that there are only three systems on the token ring (IBM and HP are the other two), none of which are being added or removed from the ring, and only one of which report the ring recoveries, namely the Alpha. We are working to get the additional information you requested, like the netstat -s. Meanwhile, may we confirm whether the sole cause of the "ring recovery" messages is supposed to be nodes physically being removed or added to the ring, please? I will also mention that the hardware has been replaced, resulting in no changes to the symptoms. Nancy Flavell Digital UNIX Network Support Specialist Customer Support Center, Colorado [Posted by WWW Notes gateway] | |||||
| 8760.5 | SMURF::GILLUM | Kirt Gillum | Fri Mar 21 1997 17:04 | 17 | |
The reason that none of the other systems display the message is
because they probably ignore it. I think I'll change the driver to
ignore the event. It would save customers from getting concerned.
From the TI TMS380 Second Generation Token Ring User's Guide...
RING_RECOVERY: This bit is set to one when the adapter observes claim
token MAC frames on the ring. The adapter may be transmitting the
claim token frames. This bit is reset when a ring purge frame is
received or transmitted.
So the ring monitor probably has detected an error condition and sent
around a claim frame. Typically this happens when a node joins or
exits the ring, but can also occur for several other reasons (like not
seeing the token within the token rotation time).
| |||||