Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

Note: I am not a network engineer. My background is coding but I'm just helping someone out with a small hardware upgrade.

Basically, we're talking about a small network of 14 computers currently running on a bog-standard 100Mbit 3Com switch. We've inherited a second-hand (although never used) Dell PowerConnect 6224 L3 switch which supports PoE and runs at 1Gbit and we'd like to take out the existing switch and put the new one in.

The 14 computers are actually thin clients which boot from an LTSP server via PXE.

I assumed (apparently incorrectly) that this would be as simple as unplugging everything from one switch and plugging it all into the other switch. However, when I do this, the thin clients only seem to be able to successfully PXE boot some of the time. The rest of the time the boot process just seems to hang. This literally never happens with the L2 switch. My only conclusion is that this L3 switch is doing more than what the L2 switch does and that the switch needs configuring before it can be used. (I've read information about the differences between L2 and L3 but I don't really understand it all)

I have had a look at the docs on the Dell website but the config language looks incredibly complicated.

So, my questions are basically: Am I out of my depth here? Is it silly to assume that this switch can be a simple solution to upgrade the network from 100Mbit to 1Gbit? Do I need an expert to configure the router or is the learning curve not that steep? Perhaps there's a "dumb mode" I can activate on the router somehow to make it work like a regular switch without any of the fancy features it probably has?

share|improve this question
1  
You should upgrade that switch to the latest firmware as soon as possible. The firmware updates fix MAJOR issues, though since you're only doing L2 it may not affect you as much. Still, I'd do it. –  SpacemanSpiff Aug 29 '11 at 13:37

4 Answers 4

Unless you specifically configure a layer 3 switch to do some customized actions it should behave by default as close as possible to a layer 2 switch, and it also has some added perks on top of that so it's always an interesting addition.

The kind of problem you're having looks to me more like one of these problems:

  • Cabling not prepared for 1Gbit: Sometimes even if the cabling says so its not prepared for 1Gbit, the best way to test this is to either use a hardware tester or perform some basic network testing from each network point, if you keep losing packets that's your culprit.
  • Switch not negotiating properly with computers: This is also sometimes a culprit of cabling, or of very cheap network infrastructure, sometimes the switch (L3) and the network card don't agree on the negotiation, so one half ends in half duplex while the other half ends in full duplex, this would show up as intermitent connection drops, if that's the case it should show up in the switch logs. The best way to solve this is to force the switch to negotiate specifically at one speed at duplex state, this is done in my Dell L3 switch with the following commands:

    speed 1000 duplex full

Let me know which one works for you! And good luck :)

share|improve this answer
    
Ok, thank you for these suggestions. I won't be at the site again for another week but I'll try what you've said and then report back here on my progress. –  Joseph Aug 29 '11 at 12:47
    
+1 for cabling, I cant tell you how often cheap cabling causes things like this. –  ErnieTheGeek Aug 29 '11 at 13:25

As lynxman stated in his answer, an unconfigured L3 switch should behave as a L2 switch. As womble stated in his answer, a second-hand switch may have a configuration from the previous owner. Here are my thoughts on the problem:

  1. Reset the switch to factory default. This will ensure that any previous configuration is wiped out.

  2. The problem you're seeing with the PXE clients is likely due to Spanning Tree Protocol being enabled on all ports. You should enable spanning-tree portfast on those ports, which can be done from the CLI or the web interface.

share|improve this answer

Second hand switches can often have the previous configuration left on them, in which case it could be doing just about anything. Dig out the manual, and work out how to reset the switch to factory defaults.

share|improve this answer
    
Sorry, I should have been more clear. It's "second hand" as in it had a previous owner. However, it was still sealed in the original packaging when we got it and has definitely never been used before. I'll update the original question with this clarification too. –  Joseph Aug 29 '11 at 12:50

This is caused by spanning tree. The switch is making sure that the new link doesn't form a loop in your network before it risks forwarding traffic on it.

When a port is first enabled, the switch needs to interrogate what is connected to that port. If it's another switch, or another port on this same switch, a decision has to be made whether to forward traffic on that port or to block it. This takes about 30 seconds, and likely the station is giving up on PXE faster than that.

The fix assuming you know this link will never, ever form a loop in your network is to enable "fast link" or "spanning tree portfast" on the port.

On the ports that you will only connect to stations:

interface ethernet 1/gX
spanning-tree portfast
share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.