Spanning tree (for bridge)?????
 

I am having trouble with two machine running RedHat 8.0. Each machine has two NICs configured by DHCP on my modem/router. The host names on these machines that a pingable from Win2K are bound to eth1.

The problems are that eth0 is flooding the LAN with "Spanning Tree (for bridge)" messages and the eth0 on both machines keep broadcasting NetBios name requests. The two Win2K machines on the LAN also have two NICs each but are relatively quiet.

Can anyone suggest what is going on I how to fix it?

You put two network cards into one computer and plugged them into the same network? That doen't double your bandwidth but it does screw up ethernet.

Spanning tree is a router protocol that (in theory) fixes it if you make a mistake in your networking, like oh I dunno say plugging two network cards into the same network.

Solution: one network card per machine.

edit: Baldy is correct (again) spanning tree is a layer 2 aka bridge protocol and a bridge loop is the mistake I was referring to (forgot what it was called) But yes you can set up linux as a bridge, I belive it involves a kernel recompile and I wasn't aware that it did STP either.

First things first, disconnect the second network card.

Secondly why are you trying to do this?

Third, if you have samba running then netbios name queries are normal so don't worry about it.

Fourth, the spanning tree issue a bit more difficult. Redhat should not be transmitting Spanning Tree packets unless it thinks it is a Spanning Tree Bridge. How have you got the little beasty configured?

STP is a protocol used by bridges, not routers :), and is used to identify and shutdown bridge loops. Are you sure its the Redhat box I didn't know linux supported bridging and STP, but then I'm hardly a Linux expert. I have done a lot of scoping on my little network and I've never seen SuSe chucking out STP. You may have other network problems, depending on your switch topology, give us a description.

Mr Pedantic here... :)

>You put two network cards into one computer and plugged them into the same network? That doen't double your bandwidth but it does screw up ethernet.

This isn't strictly true either. You *can* do exactly that, and using something called 'port aggregation' get exactly twice the bandwidth with two cards on the same subnet.

:P

See the following link :

http://www.beowulf-underground.org/c...&mode=detailed

I'm currently doing just that, although I'm using a proprietary 4-port network card with vendor-specific software driving it.. but um.. it's cool stuff :)

Slick.

I don't wanna.
Cause I got a pair of Linksys NICs with the last two switches I bought and thought I would be able to control what went where. Cisco VPN Client software on one Win2K machine is an additional complication. The new version modifies TCP/IP settings on both adpaters in a Win2K machine including mucking around with DNS. The previous version seemed to only grab one NIC abd so non-VPN traffic seemed to me going to the other adapter.
Then why don't the two Apple computers running Samba broadcast like this? For that matter, why don't the two Win2K machines (isn't Windows the root of all samba evil?) broadcasting to this extent?
In terms of protocols, etc. - pretty much like all the other adapters - DCHP, get DNS from "provider" which, to RedHat seems to mean the router. Tthe router gets DNS servers dynamically from my DSL ISP. In terms of "topology", it gets a little complicated. I have tried to ensure that the NICs in machines with two are connected to separate LAN segments that uplink to the switch that is the uplink to the router. I may have mislabeled or misconnected the second NIC on the machine that is "spanning."

I have run the LAN with four machines with two NICs and other machines with one for months without these problems. Lots has changed - OS updates, rewiring the "topology", new VPN client software. I don't think the problem is due to simply having two NICs in some machines.

Um.. oky. Darin was right. :)

The problem is *entirely* because you have two nics in the machines. And more specifically, because you have two nics which are connected to the same subnet.

Unless you're doing that port aggregation stuff above (which you're not) you can't do this.

Sambra broadcasts are out of the topic of this thread, and I went into it in quite some detail in some other thread somewhere (It has Samba in the title, I forget what).

Seriously. It's not going to work. I'm having great difficulty in understanding your topology, but from what I *do* understand, you're going to have to re-subnet one card, or remove it.

Slick.

I had redone all the cable connections recently and did find a wire that was not plugged in where I intended. When I fixed it so that all eth0s went to one switch and all eth1s went to another, the "spanning" thing went away. This also re-established essentially the same wiring that I had with earlier versions of RedHat which didn't generate all the other broadcast traffic.

Let's see, this is the networking forum, samba provides network services for file and printer sharing, I started the thread, and my initial post mentioned samba. Why don't you get that bug removed before it crawls up in there any farther?

Against my better judgement i'm going to post another reply to this topic. You don't seem to want to follow any of the advice posted here, not even as a troubleshooting step, so I really don't see the point in offering more advice.

For simplicity I stated that you could not plug two NICs into the same network, but I concede that slick is correct, you *can* set up two network cards using port aggregation also known as port trunking, adapter teaming and probably many other buzzwords but unless you *do* set them up this way they are not going to double your bandwidth.

It is also possible to actually connect two network cards from one machine into one network and not "see" any problems, but this may be due to just the simplicity of the computer's network stack; Notice how windows didn't give a rat's arse and the apple was too nice to complain but linux saw that something was wrong and threw a fit.

I'm sensing here that people are offering suggestions and you don't wish to follow the advice...
My suggestion again is to set up your network so that each computer only has one network card and see if the issues go away.

From there, if you truly wish to have multiple network cards in each machine for speed or redundancy or whatever I suggest you look into how to properly set this up:

http://www.beowulf-underground.org/...0&mode=detailed
http://www.intel.com/support/network...ns/teaming.htm

Darin,

I've been having trouble getting a reply posted. The first thing you take as resistance to advice was an attempt at humor. The second was in response to Slick's remark about being off topic. I do appreciate your response and do want to learn what is going on.

We seemed to agree that the remaining problems are due to differences in the responses of Linux and other OSes I am running when faced with two NICs the same subnet. These problems did not occur until I went from RH 7.2/7.3 to 8.0.

The machines in question are big towers and jammed under tables so physically removing the second NICs will have to wait until the weekend. I did try disabling the on-board NICs in BIOS on them. This was fine with Win2K but RH 8.0 somehow reactivated the NIC and all hell proceeded to break loose on the LAN. Peace was eventually restored by getting things back the way the were - resetting BIOS to activate the NICs again.

This actually is off-topic but RH 8.0 also seems to think the sound card in this machine has been removed almost every time Irestart after a power down. The on-board sound chip is disabled in BIOS and there is a PCI SB Live. Weird.

I can't get to the first link you posted but the Intel stuff was accessible and looks interesting. We'll see want happens when I drag the beasties out and remove the extra NICs.