norm@cfctech.UUCP (Norm Meluch) (09/28/89)
Greetings fellow netters,
I had the following problem the other morning and I was wondering
(as do most people who post to the net) if anyone as run across a similar
problem, and perhaps even had an explaination.
Our setup here at CFC is an AT&T STARLAN 10 network.
Arranged roughly as follows:
ALPHA * BETA
*
+-------+ +---------+ * +---------+ +-------+
|10 x PC|>--+--<|10 x PS/2| * |10 x PS/2|>--+--<|10 x PC|
| Client| | | Client | * | Client | | | Client|
+-------+ | +---------+ * +---------+ | +-------+
| * |
+--------+ | +---------+ +------------+ +---------+ | +--------+
|386 UNIX|>-+--<|15 x Hubs|>--<|10:10 Bridge|>--<|15 x Hubs|>--+-<|386 UNIX|
|Mgr/Srvr| | +---------+ +------------+ +---------+ | |Mgr/Srvr|
+--------+ | * | +--------+
| * |
+--------+ | +-----------+ * +-----------+ | +--------+
|386 UNIX|>-+<|5 x 3B2/600| * |5 x 3B2/600|>+-<|386 UNIX|
| Server | +-----------+ * +-----------+ | Server |
+--------+ * +--------+
AT&T UNIX SysV R3.2.1 everywhere.
AT&T Net Support Util 1.2 everywhere.
AT&T NAU driver 3.2 for 3B2s.
AT&T OSI Net Program 3.2 everywhere.
AT&T StarGROUP 3.2a for servers everywhere.
AT&T Client Program 3.1 for PCs PS/2 or AT.
The problem occured the other morning when all of the sudden the
managers on both sides of the bridge show a couple of the PC clients, a 386
server, and one of the manager's server cards go down!
After rebooting several machines, checking wiring, making sure power
was ok, etc etc etc. Only one central item was left to link together all of
the machines that went down.
All of the machines were on the "BETA" side of the bridge, and all of
the nodes that would not function contained network cards for a AT style bus.
(mfgr Ungermann/Bass).
A closer check of the network statistics on the PC cards showed a
*serious* number of CRC and alignment errors. It is interesting to note that
the all network cards including microchannel, 3B2, and PC cards showed the
errors. The PC cards were the only ones that couldn't recover.
At this point I shut down the bridge, and powered off all unnecessary
hubs, and called AT&T. This (doubtless the phone call was the true fix)
seemed to stablize the network activity. After a long phone consultation
AT&T could say nothing about what the cause of the problem was. There
were no errors in any log at all! (Why should there be for a hardware problem).
Later all pieces of the network were powered on again and every thing has
been fine ever since.
My questions are:
1. Is there any way to know what was causing the network errors?
2. How can I try to isolate the problem quickly the next time
(and you know it will) occur.
3. If you ever had this problem what was the cause?
Here is a copy of the network statistics from a PC node during the problem.
Network Statistics for G061B.ADM.CFC
Jumper Settings: I/O @ 360, IRQ 3, Duration: 16 days,
RAM address: e0000 13 hr, 57 m
Software Version: 3.1 Active Sessions: 4
Physical Address: 08006a024423 Frames Transmitted: 160126
Max Frame Size: 1072 Frames Received: 217950
Max Sessions: 32 Bytes Transmitted: 8654469
Bytes Received: 11541928
NAMES ATTRIBUTES
1. G061B.ADM.CFC Unique Collisions: 56646
2. G061B.ADM.CFC Unique Retransmissions: 164
3. @G061B Unique VC Connections Refused: 0
4. SBS000000. Unique Frames Aborted: 1304
CRC Errors: 226
Resource Errors: 73
Alignment Errors: 4096
Thanks for putting up with this long article.
Please respond by e-mail or net. I care not. Other Sys admins might though.
:-}
- Norm
|~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Norman J. Meluch ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~|
| Mail:{sharkey|mailrus}!cfctech!norm Fax:(313)948-4975 Voice:(313)948-4809 |
|~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~|
| Note: The opinions expressed here are in no way to be confused with valid |
|_______ideas or corporate policy.____________________________________________|norm@cfctech.UUCP (Norm Meluch) (09/28/89)
In article <16387@cfctech.UUCP> norm@cfctech.UUCP (Norm Meluch) writes: >(mfgr Ungermann/Bass). Oops. The microchannel cards are made by them. The PC cards have no mfgr. stamp and could be an AT&T original. I do not know. - Norm |~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Norman J. Meluch ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~| | Mail:{sharkey|mailrus}!cfctech!norm Fax:(313)948-4975 Voice:(313)948-4809 | |~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~| | Note: The opinions expressed here are in no way to be confused with valid | |_______ideas or corporate policy.____________________________________________|
kratz@bnrgate.bnr.ca (Geoff Kratz) (09/30/89)
What type of media are you on? UTP, thin coax? We are using
UTP as the physical for ethernet, and clock specs on the controllers have to
be very VERY tight. Maybe you have something with an out of spec
clock?
--
Geoff Kratz Bell-Northern Research, Ltd. Ph: (613) 763-5784
Internet Systems P.O. Box 3511, Station C FAX:(613) 763-3283
Ottawa Ontario Canada K1Y 4H7
kratz@bnrgate.bnr.canorm@cfctech.UUCP (Norm Meluch) (10/04/89)
Sorry about that. We're running AT&T STARLAN 10 on UTP. - Norm.