And to ask you who else has come across the danger to corporate networks posed by the convenience of auto-crossover ports.
I had this at a middle school I do work for, One little darling student did exactly this.
<snip>
It is annoying that such a thing can happen, but to be honest it's no worse than in the old days of a missing t-piece or a cable fault, bringing down a whole coax lan.
True, those were the days. But the Base-T system has been so comparatively rock-solid reliable and proof against all the problems we used to have that it has been a shock to find that it is so easily compromised. And the problem has only arisen with the more recent equipment!
You should be using spanning tree protocol (sometimes called spamming tree protocol when it breaks) which should allow these kind of redundant links that you are creating. Indeed redundant links should be/are a good thing.
Yes, redundant links are, I am building up a number, and I have looked at STP. The main reason I haven't used it is that I prefer the idea of subnets, routed, RIP and routers instead: routers based on multi-port FreeBSD boxes. I don't see how STP (or subnets) cure this switch-jamming problem, except that the more intelligent the device the more capable it is of preventing the jam from spreading back onto the network spine. However, we have some quite large (multi-building) branches containing hundreds of outlets that are all within the same easily-jammed area.
Luckily I don't yet have a large number of auto MDI/MDIX switches and am hoping to obtain more older models, which are immune to this problem because normal users don't possess crossover leads.
On Sat, 29 Nov 2003, Christopher Dawkins wrote:
Yes, redundant links are, I am building up a number, and I have looked at STP. The main reason I haven't used it is that I prefer the idea of subnets, routed, RIP and routers instead: routers based on multi-port FreeBSD boxes. I don't see how STP (or subnets) cure this switch-jamming problem, except that the more intelligent the device the more capable it is of preventing the jam from spreading back onto the network spine. However, we have some quite large (multi-building) branches containing hundreds of outlets that are all within the same easily-jammed area.
Hi,
Spanning Tree is a level 2 protocol where as RIP is a level 3 protocol. Ie Spanning tree works at MAC address(Ethernet), RIP works on IP address(TCP/IP level).
A switch in a Spanning Tree network learns the MAC address of other devices attached. If it detects a loop, it will automatically block one of the lines that are feeding into it. A master hub has to be defined in the system. This is done by setting a prioritory level ( believe it's between 0-65535). If all switches are set to the same priority it will automatically work out which device has the lowest mac address and use that as master.
For example
____ | 1 | |____| / \ / \ ___ / \ ___ | 2 |____________| 3 | |___| |___|
Hub 3 has two alternate paths to talk to hub one, either directly or via hub 2. This would cause a loop situation, as you are already aware. Using Spanning Tree, Hub 3 would automatically block all data on either the port that connects it to hub 1 or hub 2. This would depend on the spanning tree prioritory set in each of the hubs. If Hub 1 had a prioritory of 1, hub 3 would block all data from hub 2, so all data would go via hub 1. If hub 1 were to fail, the block would be automatically removed.
That is how I remember it working when I did the course in configuring Foundry Routers and switches a few months ago.
Hope that helps
Chris
Chris *************************************************************************** E Mail Chris@glovercc.clara.co.uk WWW http://www.glovercc.clara.co.uk Someday, we'll look back on this, laugh nervously and change the subject. -Anon
On 2003-11-29 22:51:04 +0000 Chris Glover chris@glovercc.clara.co.uk wrote:
That is how I remember it working when I did the course in configuring Foundry Routers and switches a few months ago.
Is it normal to dance around the room when there's a Chris Glover ALUG post? That was one of the best explanations of STP I've ever read. (I hope it was accurate. I am not a networking technology expert. :D)
I think part of the reason that failing STP has the nickname "spamming tree protocol" is that "stateful" connections like TCP can throw out the repeated packets, but "stateless" UDP ones can't. Generally, resolving hostnames into IP address numbers is done by UDP, so a spamming tree often kills local DNS servers early on, as new requests get added to the looping old requests and they don't normally spot the repetition. The DNS server responses are also looping, too, adding to the traffic.
On Sun, 30 Nov 2003, MJ Ray wrote:
On 2003-11-29 22:51:04 +0000 Chris Glover chris@glovercc.clara.co.uk wrote:
That is how I remember it working when I did the course in configuring Foundry Routers and switches a few months ago.
Is it normal to dance around the room when there's a Chris Glover ALUG post? That was one of the best explanations of STP I've ever read. (I hope it was accurate. I am not a networking technology expert. :D)
Yep! :-)
I may be in deepest darkest South Essex, but I do still read this list.
As for accuracy.. I hope so :-)
Chris *************************************************************************** E Mail Chris@glovercc.clara.co.uk WWW http://www.glovercc.clara.co.uk Someday, we'll look back on this, laugh nervously and change the subject. -Anon
-
Chris Glover -
Christopher Dawkins -
MJ Ray