Sun Managers,
[I got several more relevant responses after posting the SUMMARY for this, plus
a number of `me too' replies with a request for the responses I got, so I'm
reposting this with the new stuff included. In particular, the last response
from D.B. Tosic (dt@ffm.deu.ubn.net) looks like it might be relevant to our
problem, especially since I haven`t been able to find any cable anomalies yet.]
On Monday, June 20, 1994, I wrote:
> On the console of our 4/490 running 4.1.3_U1, we often see the messages
> `ie0: Ethernet jammed' usually followed by multiple occurances of 'ie0: no
> carrier'. There doesn't seem to be any actual problem associated with
> these messages - as far as I can tell, there's no real performance
> degradation occuring that the messages suggest. I installed the ie0
> Ethernet jumbo patch (100570-05), which according to the patch description
> sounded like it would eliminate the messages, but it had no apparent effect
> and we are still seeing them.
>
> Anyone out there familiar with this problem ? Any suggestions would be
> much appreciated. Thanks in advance,
I got responses from the following helpful persons:
Mr T Crummey (DIJ) <tom@sees.bangor.ac.uk>
john@pendragon.zim.bms.com (John Turner)
Kai Birger Nielsen <bnielsen@daimi.aau.dk>
dmaccara@bank-banque-canada.ca (Donn MacCara)
raoul@MIT.EDU (Nico Garcia)
perryh@pluto.rain.com (Perry Hutchison)
dave@gaspra.pd.com (David Gilinsky)
SAMPSO01@UTSW.SWMED.EDU (Mike)
ric@Artisoft.COM (Ric Anderson)
hedstrom@inf.ethz.ch (Leif Hedstrom)
dt@ffm.deu.ubn.net ( D.B.Tosic )
Generally, the consensus is that we've got a cabling problem of some sort,
either too much length or a bad connector. I'm sure it's not a length
problem in this case, but a bad connector is always a possibility, so I'm
looking for that.
Thanks to all responders. Response text is attached.
-- David Way McDonald Observatory/Astronomy Dept.- Univ. of Texas, Austin (office) RLM 16.206 (voice) 471-7439 (internet) dpw@astro.as.utexas.edu -- >From bnielsen@daimi.aau.dk Tue Jun 21 01:42:53 1994We have seen the same thing here. Older sun stations 3/280 and 4/280 complain about this when the net is under heavy load. Problem is that "heavy load" in their terms is not really that unusual anymore. I think we got it pinpointed to the ethernet controller chip. Details can be found if this is not enough. I can trigger this with two sparc stations talking together on the same net segment as the 280's. My present interpretation is: Either serious netproblem (which will be indicated by lots of other machines too) or "I tried to send a packet 16 times and couldn't.". The last one is what you would expect if two machines are taking the full bandwidth for even a short period. -- >From tom@sees.bangor.ac.uk Tue Jun 21 04:06:00 1994
Hello David,
This message usually means there is a loose ethernet cable somewhere and you are getting intermittent breaks in the ethernet continuity. I've never heard of spurious versions of this message. In my experience there has always been a hardware problem somewhere on the network. Check all the cables connected to that segment of ether and make sure no-one is unplugging anything!! During these periods, you should find that the netowrk ceases to operate. -- >From john@pendragon.zim.bms.com Tue Jun 21 08:44:18 1994
Way back in my youth :-) we had a 3/280 that had a messed up slide lock on it ie0. Every so often the cable would get bumped, and we would get a whole bunch of "no carrier" and "ethernet jammed" messages. Carefully reseating the cable solved the problem. Of course, when we got these messages, the machine was unreachable by anyone on the network. -- >From dmaccara@bank-banque-canada.ca Tue Jun 21 09:20:23 1994
Hi, This won't be of much help, but we have had the same experience on a Sun 4/470, including the installation of the patch. My understanding is:
This is a message specific to the Intel Ethernet driver (ie0) so that you would never get the message on a SPARCstation (le0), for example
Although we have 4 servers with device ie0, the error only appears on one subnet and it is the busiest one by far. However, the network can be very busy sometimes and we never see it.
In the past, the message has occurred when we were having problems in the wiring closets. In these cases, though, we would get hundreds of messages and several workstations would appear hung.
Now that the wiring problems have been fixed, I tend to ignore the occasional "jammed" message. -- >From raoul@MIT.EDU Tue Jun 21 09:41:20 1994
Hmm. Double check for cabling problems: I saw something like this when a !@#$ Mac user mis-terminated the Thinwire at his node, effectively double-terminating it at that end. -- >From pluto!perryh@qiclab.scn.rain.com Tue Jun 21 04:02:52 1994
It could be caused by a bad transceiver, drop cable, or interface; or by a damaged backbone cable (thicknet is more fragile than it looks -- the insulation between the shield and the center conductor can get mashed); but the one time I ran into it the cause was a network segment badly over its maximum specified length. (This particular one was thinnet and something like 400 meters long. I believe the specified maximum for thinnet is 200 meters.) The messages went away after a repeater was installed to split the segment into pieces of acceptable length. -- >From SAMPSO01@UTSW.SWMED.EDU Tue Jun 21 11:40:17 1994
I've had quite a few of those error myself, and they have ALWAYS been attributable to some weakness somewhere in the cabling. Either bad terminators, loose but still conductive connections, or cable damage of some sort. This is on a smallish thinnet network. Hope this helps. -- >From dave@gaspra.pd.com Tue Jun 21 11:10:21 1994
We've got a 4/110 here running SunOS 4.1.3. The only time I've seen the message you described, we had a bad stretch of coax cable between two of the machines on our local net. -- >From ric@Artisoft.COM Wed Jun 22 23:39:40 1994
Sorry I missed the original post on this. A 4/490 at the last place I worked exhibited this symptom during periods of peak net activity.
Once we installed an etherswitch to segment the net, this problem went away. Might be time to get out the ethernet monitor and see what the traffic volume looks like (if your connector is ok) -- >From hedstrom@inf.ethz.ch Thu Jun 23 03:14:34 1994
I didn't see you original posting, but I have had similar problem. For me, it was *not* a cabling or connector problem. Did you see any message like "mbuf table full" or something like that?
For me (on a Sun490) it was another machine spraying our server with tons of packages. A reboot of the "client" solved the problem on the server. It has happend only two times for me, but the effect on the server is that it is completely "stuck".
The "mbuf" stuff are some buffers used by the TCP/IP driver (I think...) -- >From dt@ffm.deu.ubn.net Thu Jun 23 06:04:39 1994
Regarding your question here is personal experience: Last year I had to go to customer site and troubleshoot a same problem as you have (intermittent ie0 no carrier).Following was found out: 1.Measurement with HP Lan-analyzer ==>Each time a network load reached 40-50 % message ie0 no carrier will appear.It was one 4/490 with approx. 100 workstations connected.Actually there were two segments with repeater in a between.After the repeater was replaced with Retix-Bridge message will appear but only 2-3 times a day (before that there were hundreds). Definetely a message was corelated to network load. 2.Network jumbo patch was installed but it DID NOT HELP !(Check with your local Sun rep if they have any new ones). 3.Message will appear only with ie0 BUT NOT with le0(different Ethernet chip).
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:09:04 CDT