Ok, I have looked but didn't find anything along these lines here in the BBS. So I am sending out another network thread. I don't think this is a problem with the Empeg, but with my network. So here goes.

To make a long story short...I have a 3com Switch 10/100 that I am now using to network everything together. I have moved all my computers over to this switch. But in doing so I can't find the Empeg on the network any longer. I have narrowed it down to the component that makes the empeg be seen on the netywork. I was using a Cabletron 10base-t Hub. This hub somehow enables the Empeg to work with my network. If I leave everything on the new 3com switch and simply uplink this hub to the 3com and attach the Empeg to the Cabletron, the Empeg is there. I can set an IP on the Empeg or leave it as DHCP and either way it works. Niether setting will work without the Cabletron hub inline. My question is why must this hub be here for the Empeg to function? What is this hub doing?

Specifics on the network.
Running Win98se
1st computer is static 192.168.0.1 with ICS installed
2 other computers are set to DHCP(per ICS settings)
Using a 3com Super Stack II 3000 12x Switch 10/100(yes it does verify that when the Empeg is connect it is 10HD
I have an Apple Airport attached to provide access to my laptop(makes no difference if disconnected)
A 4th computer(laptop) uses the ICS(via the Airport)
The Empeg(when set to DHCP) shows eth0 DHCP to be 192.168.0.1 which is the ICS machine and get assigned the next free IP on the network.
The hub is a Cabletron MicroMMAC24E 10BASET-HUB with LANVIEW.

At First guess, I'd say you have a duplex issue. The empeg has to be at 10 mb, half duplex. The hub will only do this, and won't try to push it to full duplex.

The 3coms (and HP switches, as I have cursed before some firmware updates) autodetect what setting is needed. Sometimes they miss. They might get the 10/100 part, but the half/full slips by. And that can be fun to figure out sometimes.

Leaving the hub in the equation forces half/10mb.

The problem with standards is that there are so many to choose from.....

I have checked the 3com switch in the managment options. It is seeing the empeg as 10 half duplex. Unless it just reports it wrong. But it is seeing everything else as it should and reports correctly.

I have even tried a few other hubs. I have another non-intelligent Cabletron hub. I have tried a Linksys 10/100 hub. None of them allow the empeg to operate.

I used to administer some 3Com switches and I saw similar problems with 10Mbps Ethernet devices. Check the errors on the switch and on the empeg (if you can -- it will be ``netstat -i'' if that command exists on the empeg). Force the switch to 10/half and it'll probably start working. I never really researched much (they were older Macs that I just got Fast Ethernet interfaces for), but I always assumed that what was happening is that the switch's autonegotiation was going nuts. When the IEEE released the Fast Ethernet spec, one of the things included was an autonegotiation feature that would help the devices figure out whenter to work in 10/half, 10/full, 100/half, or 100/full mode. I think that the 3Com switches continue to try to autonegotiate after its already figured it out.

Silly question, but you have verified that your cable is correct? Not using a crossover post on the cabletron?

(if you are getting it to respond to the management screen, i'd have to say the wiring is correct enough)...

That's wierd. The only thing I can think of was a similar problem on our HP switches... Before upgrading their firmware, any 3com card capable of 100Mbs operation would perform like [censored]. Max throughput of about 1-2 k per second. Even if you forced both the card and the switch into 10hd, it tried to run 100FD. Put a 100mbs switch (cheap, $60 variety) between, and it would operate 100mbs no problem, getting 7000Kps no problem. Of course, the majority of these were in Dell Optiplexes, integrated. Pain in the ass to replace an integrated net card.

Found an updated firmware about 2 weeks later, and applied it to the HP's... Everything has been cheery since.

Not saying this is your problem, as you've tried different hubs/switches... but it does seem VERY strange that the empeg ONLY works on that hub.

Have you tried putting both the empeg and your PC on the linksys 10/100, taking the 3com out of the equation?

Ok, I changed the 3com switch to 10hd only, on that port. Still no go.

Yes, I have tried the Linksys 10/100 switch. It won't work alone. As I have a few more devices from time to time that we hook up, I needed more ports than what I have on the Linksys 4-5(5 being uplink). So I was using the Cabletron with all my 10baseT and uplinking to the Linksys. I have been using that for the Empeg all this time. Now that I have an 10/100 autosensing switch with more ports I want to replace the Linksys and the cabletron with the 3com.

P.S. I am using a cross-over cable from the Cabletron to the 3com as neither one has a dedicated uplink port. I don't see a problem with that, do you?

Just to verify that we're working in the right direction, can you see any errors on that port, either from the empeg or switch side?

Also, it weird, but it appears as if 3Com has removed almost all refernces to the SSII 3000 from their web site. The only place I can find it is some apparently leftover stuff from their old (blue, before the interlocked rings) web site. It's almost like they disavowed any knowledge of that switch. Which is not totally amazing, as memory of the problems I had with it come flooding back. You might want to see if you can update the firmware. The only page I found is this old SSII 3000 Software Library page.

You might also want to try forcing the switch to full duplex. I doubt that the empeg is full-duplex, but I don't have any way to test here, and it can't hurt any more than it's hurting now.

OK, with the Cabletron in line and on port 1 of the 3com, I can sync with the Empeg. About 5-20 collisions reported, and 1-8 errors. With the Empeg straight to the 3com, Emplode never finds the unit. After about 5min there is about 3000 collisions and maybe one error reported on that port.

Now to try 10 Full Duplex, It actually finds the Empeg now. We are making progress. But it fails everytime while on the "Retrieving Database" task, with a connection timed out. The switch reports no collisions, but about 25 errors.

I have searched the 3com site as well and have updated to the lastest 3.23 version in the switch.

Now what do you think?

I would not guess that it is the root or a big part of your problem, but one potential confounder to consider as you troubleshoot is the ARP table on the 3Com. I think the SSII is smart enough to cache ARP entries associated to port/s, and if you are moving things like the Empeg around to different ports as you troubleshoot, the persistence of stale entries may add to unpredictable behavior that makes your troubleshooting more frustrating. There should be a way to clear ARP on the 3Com so that it forces broadcasts to refresh entries for Empeg, PC, etc.

Looking back at an earlier post, I see something I don't understand (apologies if someone already covered this) and have a few random questions:

1st computer is static 192.168.0.1 with ICS installed
(.....)
TheEmpeg(when set to DHCP) shows eth0 DHCP to be 192.168.0.1 which is the ICS machine and get assigned the next free IP on the network.

This last statement makes it sound as if the Empeg ultimately gets the same address assigned as the "1st computer running ICS". Am I reading that incorrectly? [edit: naw, I re-read that and see what you are saying]

What subnet mask is specified for these hosts? Have you tried all of this with a static IP/mask on the Empeg and on the PC running Emplode with only those 2 hosts on the network?

Also, it sounds like all link status lights on Linksys, 3Com, etc. are lighting up the way you expect, true?

Well, to the ARP tables question. I can take everything off the 3com but my computer and the Empeg. I can reset the 3com. I would think that would clear the tables.

Just to clarify on the issue of the Empeg reporting eth0 DHCP is 192.168.0.1, it also reports eth0 IP as 192.168.0.2(the next available IP address. I am assuming here that the ICS machine(192.168.0.1) is serving the Empeg that IP.

Now the above only works if the Empeg is connected to the Cabletron hub then uplinked to the 3com switch. Or if the Cabletron hub is the only device used.

Everything is using the same subnet. 255.255.255.0

Yes, all the lights on the other hubs light up for 10 or 100 depending on the device connected. Even the 3com switch reports each port correctly even assigning Full Duplex and Half Duplex.

So basically, all my equipment works on everything but the Empeg. If I place the Cabletron in line with the Empeg. The Empeg is now seen on the network(static or dynamic.)

This leaves me with two questions. Why doesn't it work with anything other than the Cabletron? And what is the Cabletron doing that is making it work?

I found the problem. I have a bad cable that I have been using on the Empeg. It gives me a link light on each end. The Cabletron works with it. How could that cable be bad. I had just assumed(yes I know how much trouble that words) that if the cable worked with one it would work with all.

Now I remember reading about in the Cabletron manual that it will work with miswired cables. It just corrects the problem internally. They recommend you fix the problem and not keep using miswired cables. My problem with that is, if it worked and gave link lights and tranferred data, how would you ever know that you had a miswired cable.

So basically to answer my earlier question, "What is the Cabletron doing?" It is correcting a bad cable problem.

So thanks to all the people that worked on this for me.

P.S. I cut that cable up and threw it away!!!

94cobra: So basically, all my equipment works on everything but the Empeg. If I place the Cabletron in line with the Empeg. The Empeg is now seen on the network(static or dynamic.)

Well. I don't have any particularly bright answers. Other folks' earlier replies are probably closer to the mark -- a subtle layer 1/2 duplex/framing problem b/w Empeg and 3Com. In your shoes, I'd want to fight with the 3Com until I won and I'd keep it a 2-node static network while fighting with it for troubleshooting purposes (if you can't get it to work like that, then...). It's complete voodoo, but aside from swapping new cables and such, I'd even try a different network, say 192.168.1.0; this doesn't really seem like a layer 3 problem, and 192.168.0.0/255.255.255.0 should yield usable host addressing, but different vendors' equipment sometimes have funny ideas about what constitutes a "zero" network, and it's probably at least worth ruling that out.

I have a bad cable that I have been using on the Empeg. It gives me a link light on each end. The Cabletron works with it. How could that cable be bad.


Glad you found your problem.... Cables CAN give the link indicator lights, but still be wired wrong. You've got four wires that have to work to get a successful connection. Only 2 are required to get a link light.

I'm just curious how the Cabletron "Fixes" this. And for that matter, to refresh something you had said... WHY?

If my cable is bad, I don't want it to work. PERIOD. That means that I can replace it. Having the hub rewire internally is just some crap I don't want.

Anyways, congratulations, and make sure that cable gets tossed... Even cut in two, you might decide to try to re-use... I've had tech's that have done that... only to discover the problem was about 5 inchs from the plug....

Out of curiosity, what does the 3Com say now about the duplex of the empeg link? Someone else might come along with a similar problem and since you're already looking ....

94cobra: I have a bad cable that I have been using on the Empeg. It gives me a link light on each end. The Cabletron works with it. How could that cable be bad.

Hooray. It's so nice when it comes down to the simple things. As to "how" I wouldn't stress over it too much. Aside from non-compliant or mis-terminated cables, it doesn't take much for a cable (or a port) to degrade enough to cause a problem like this.

synergy: Glad you found your problem.... Cables CAN give the link indicator lights, but still be wired wrong. You've got four wires that have to work to get a successful connection. Only 2 are required to get a link light.
I'm just curious how the Cabletron "Fixes" this. And for that matter, to refresh something you had said... WHY?


Well, IIRC, you could run 10Base-T over 2-pair Cat 3, and even fast ethernet over 1 version of 2-pair Cat 5, so maybe the Cabletron is more lenient to pre-Cat 5 10Base-T and allowed the maybe-not-wired-to-4-pair-EIA-Cat-5 cable to work where the 3Com wouldn't.

If my cable is bad, I don't want it to work. PERIOD. That means that I can replace it. Having the hub rewire internally is just some crap I don't want.
Agreed with a bullet. I saw another thread here about somebody getting a patch cable wired up at Radio Shack. I have to guess that marginal cables -- and the false economy of building cables in-house to "save money" -- are the cause of more non-productive lost hours of IT support time than, well, lots of stuff. Buy cables from places that make jillions and if they start to seem dubious, throw them away.

Well, IIRC, you could run 10Base-T over 2-pair Cat 3, and even fast ethernet over 1 version of 2-pair Cat 5, so maybe the Cabletron is more lenient to pre-Cat 5 10Base-T and allowed the maybe-not-wired-to-4-pair-EIA-Cat-5 cable to work where the 3Com wouldn't.


You can run 100mbs over cat 3 if the conditions are right. Both 10 and 100 only use 2 pair. One pair is positive/negative transmit, one positive/negative receive. The differences in cat3 and cat5 are in the twisting, predominately. The other two pair can be used for voice, if you wish... Not much reason to do so, now, but it can be done.

The fun thing you can do with 10 and 100baseT is just hooking up the receive pair. Then you can use that cable as a damn near undetectable sniffer with a promiscious ethercard... No other machine on the net is going to be able to pick it up.... which the right software can do to a normal sniffer. Not a HELL of a lot of use, but it has came in handy once here.

10mbs will run on just about anything. I've got a plant, that I inherited mind you, that has 50% or more of it's network running on 2 pair of cable, loose, unjacketed, no twist... nada. It's fuglin uckly. It's also getting rewired, but it does, surprisingly, work as is.

My guess is that the cable had a pin off, or was a crossover cable, and the cabletron detected and corrected. How, or why, I don't even want to know.

and the false economy of building cables in-house to "save money"

The Funny thing is... I have a hard time agreeing with this. It's a pain in the ass to do, but I'll trust my cables a hell of a lot more than the prebuilts. I primarily use pre-builts, but whenever there is a possible cabling issue, I'll crimp my own to make sure. Now, I've got a hell of a lot of experience in doing these, and don't recommend just anyone doing it, but......

Now, thinking about it, I don't trust anyone else's in-house cables, so maybe...

Of course, buying ANYTHING at Rat Shack is dubious....

I said: And the false economy of building cables in-house to "save money"

synergyThe Funny thing is... I have a hard time agreeing with this.

Most people who build their own cables do!

It's a pain in the ass to do, but I'll trust my cables a hell of a lot more than the prebuilts. I primarily use pre-builts, but whenever there is a possible cabling issue, I'll crimp my own to make sure. Now, I've got a hell of a lot of experience in doing these, and don't recommend just anyone doing it, but......

Actually, I honestly *can* see you trusting your own over prebuilts. What I've seen a lot of, though, is IT staff at budget-challenged institutions crimping cables during down-time out of Cat 5 leftovers thinking that they were being good citizens (of course, they didn't have a budget for a cable tester, either) and then wasting huge amounts of time later troubleshooting over these "free" patch cables. Given my crimping skills, give me a cable built by somebody who shucks 5 per minute -- maybe even with a molded strain relief or hood thrown in -- is pre-tested and that costs 4 bucks. Then if any doubt, throw it out.

Now, thinking about it, I don't trust anyone else's in-house cables, so maybe...

Yup!

You need one pair each way to get a link light at each end.

The problem being corrected by the hub may be +/- of a pair being swapped over; some chips will auto-correct this.

Hugo

+/- of a pair being swapped over; some chips will auto-correct this.

Netshop are advertising 10/100/1000 cards which will auto-correct for a cross-over cable used instead of straight-through, or vice versa. Why nobody thought of this ten years ago is beyond me...

Peter

Ah, the age-old "it was the cable" story.

If we added up all of the lost human time caused by bad cables, I think the numbers would be staggering.

Thanks for reminding me of one of the most fundamental concepts in network troubleshooting. I have added it as a bullet point in the FAQ entry on getting ethernet to work.

synergy: You can run 100mbs over cat 3 if the conditions are right.

For this you need 4-pair, don't you? 100BaseT4? Or maybe just limited distance over 2-pair???

Don't mind me. For a moment I thought your earlier post was saying you needed 4 pair, not 4 wire. More coffee... Pay attention.... More coffee....

(It really *is* nice when a problem comes down to the fixably simple versus the agonizingly mysterious!)

For this you need 4-pair, don't you? 100BaseT4? Or maybe just limited distance over 2-pair???


That's the fun with specs... Some require, some don't.

The popular one requires 2 pairs twisted. There was one that hit all 8 wires, but I don't think much became of it.

I've also found Coffee just doesn't do it anymore. Xenadrine.... Now that's a picker-up... Especially with a Dr.Pepper.

My heart is going to explode out of my chest one of these days...

synergy: The popular one requires 2 pairs twisted. There was one that hit all 8 wires, but I don't think much became of it.

That sounds like it could be the HP-backed AnyLAN VG (sp??? I actually saw it in the wild once -- a hospital that bought a turn-key lab system based on all-HP hardware). I think it would run on Cat 3 with all 8 wires. Same with 100BaseT T4 media type, I think, designed to re-use existing Cat 3 on 8 wires, but not sure how much it gets implemented in vendor equipment (or for that matter how -- can a port support both T4 and TX??).

Back to my regularly scheduled Triple Americano...

The common Fast Ethernet medium is 100BaseTX. It utilizes two pairs of Cat5 cabling and is usually terminated with an RJ-45 modular plug. The two pairs used are pins 1 and 2, and pins 3 and 6, where the numbering starts from the left when you're looking at the tip of the plug with the clip pointed up. Each pair must be twisted together. Most installations I've seen use the orange and green pairs, respectively. The other two pairs don't have to be hooked up at all, but if they are, brown is pins 7 and 8, and blue is pins 4 and 5, but the blue polarity is reversed. So a finished cable looks like:

Orange	Orange	Green	Blue	Blue	Green	Brown	Brown
White		White		White		White

When making a crossover cable, pairs one and two must be swapped, but polarity kept, on one end. I"m not sure if pairs 3 and 4 are supposed to be swapped technically, but with 100BaseTX, it doesn't make much difference, and this type of cable is hardly ever a long run, so fixing it if 4 pairs becomes required in the future is a trivial undertaking. So:

Green	Green	Orange	Blue	Blue	Orange	Brown	Brown
White		White		White		White

The main difference between category 3 and category 5 cables is the number of twists in the pairs per unit length. This translates to being able to transmit further. You can get away with using Cat3 cabling with 100BaseTX if your run is fairly short. This is easier to do with switches because they also function as repeaters, amplifying the signal, so runs can be measured from switch to switch and switch to host, as opposed to hubs, which have no amplification feature, and, as such, runs must be measured from host to host, which increases the number of different run lengths geometrically, making troubleshooting and solving individual host to host problems much more difficult.

There is also a specification for 100BaseT4, which is designed to use all four pairs on Cat3 cabling. I've seen products advertised occasionally, but seldom. Cisco switches do have modules for them, though, I believe. There's also 100BaseFX, which is run over multimode fiber.

I hope that this puts to rest all Ethernet talk. And, please, no one bring up 10Broad36.

wfaulk: I hope that this puts to rest all Ethernet talk. And, please, no one bring up 10Broad36.

But would you forget 10Base2, 10base5, or the ill-fated 10BaseAF*??

Jim

(*failed attempt to put 10Mbps ethernet over al dente fusilli)

I was reading that in the Cabletron manual about it 'fixing' an incorrectly wired cable. And thought, maybe it is a cross-over cable and I didn't know it. It wasn't though. Just not wired correctly. I didn't make the cable.

The Cabletron manual also states that you shouldn't use and incorrectly wired cable for extended time frame. My question is how would I ever know its bad if it works?

And yes I cut the cable up.

Now that you brought it up, could you explain 10Broad36?

As soon as the wife gets home with the Empeg I will look and see.

But would you forget 10Base2, 10base5, or the ill-fated 10BaseAF*??

Jim

(*failed attempt to put 10Mbps ethernet over al dente fusilli)

Just another day I saw in a local hospital reception area real 10Base5 network, with thick coax that was actually yellow run high under the ceiling, transcievers and their 'vampire' connects each in plexiglass box, and MAU cables snaking around to 286 machines running ancient Clipper app on long forgotten version of NetWare. (BTW, two topmost google hits on 10Base5 relate to a techno music band.)

Speaking of fusilli, why not, aren't they twisted enough?

I also remember when 4Mb/s ARCnet running on 90ohm IBM SNA cabling was popular. Some 'installers' did not bother to stock that cable and 50ohm 10Base2 cable, but were 'splitting the difference' and using cheap 75ohm TV antenna coax. It often sort of worked.

Ok, The 3com switch auto detects the Empeg as 10 Half Duplex. I set the switch to 10 Full Duplex and then reconnected the Empeg. Emplode would see the Empeg(so it was getting served DHCP) but it would not get past the 'Retrieving Database' task.

I think that would indicate the Empeg is only cable of Half Duplex transmission.

Just another day I saw in a local hospital reception area real 10Base5 network, with thick coax that was actually yellow run high under the ceiling, transcievers and their 'vampire' connects each in plexiglass box, and MAU cables snaking around to 286 machines running ancient Clipper app on long forgotten version of NetWare


Hell, I've GOT machines on 10base5. I've got an old VAX that we have to keep for historical reasons that is running all of it's decservers off of it.

I #(*$&@#*($&(#@ HATE Vampire taps. They are called that because it's damn near impossible to put the damn things on without drawing blood.

Around March of next year, all of the data on that VAX will be gone/no longer needed. I've already got the slugs and the shotgun is polished and ready to blow several very large holes in the thing. I'm thinking about pouring Liquid Nitrogen over it first, just for the effect.

ROFL!

Picturing the scene in Office Space...

Office Space...


Exactly.... Only with the addition of the Cryogenics.... I figure that should really make it "explode"... Slugs probably won't do well, except on the drive arrays.... On the main CPU box, buckshot with a tight choke might do better.

Rest assured, I WILL have the mini-DV camera rolling.

That box has caused me so much grief.... VMS is a good OS... When it's Fricken installed correctly. After god knows how many years of missed updates, and crappily applied 'fixes'... I shall enjoy it's destruction with great enthusiasm.

In reply to:

---

Around March of next year, all of the data on that VAX will be gone/no longer needed. I've already got the slugs and the shotgun is polished and ready to blow several very large holes in the thing. I'm thinking about pouring Liquid Nitrogen over it first, just for the effect.

---

If you're gonna do that, you could send it my way instead. We need to preserve our computing heritage, not destroy it. Or at the very least, I need to collect more crap that'll gather dust.

At the same time, I knew a guy who, about ten years ago, bought an old Intel multiprocessor machine (can't remember the manufacturer now -- Con-something?) from NC State University's computing department for cheap after they retired it and then sold tickets for people to take whacks at it with a baseball bat. As I remember, he made quite the profit.

If you're gonna do that, you could send it my way instead. We need to preserve our computing heritage, not destroy it. Or at the very least, I need to collect more crap that'll gather dust.


Oh no.... It's mine. Well, Mine and Mr. Remington. After as much blood and sweat, it will pay.

I do have a 4 Proc Unisys box that is looking for a new home... It's a beast, with a ton of memory on minicomputer hardware. And it ran Dynix, which would have been an ok OS if we could have ever gotten any of the gnu stuff to compile....

Didn't come with a compiler, wasn't available, and gcc wouldn't go. I replaced the beast with a linux box and haven't looked back.

It was pretty funny that the mission critical box ran for 3 years on 4 Pentium 60's. Brought in an Athlon and utterly destroyed it. Gotta love Moores law.

Nicely put!