mac address aging problem

nashbuff1 · ‎07-11-2006

Okay were having a mac address problem we can't quite find. Here's the whole long story.

Our setup is a frame of linux boxes all connected via bonding to 2 cisco 2950's in each frame, switch1A and switch1B. Each of the switches has an uplink for our 66 vlan and 67 vlan into our main lab switch(cisco 2948). Each linux box has 2 ports in bond0 with vlan 66 on it and 2 ports in bond1 with vlan 67 on it. To block the obvious loop, spanning tree on our switches was blocking the crossover link between the 2 switches. So both uplinks to each switch were active. Everything appeared to be working fine until...

We noticed we got duplicates alot. We traced this to mac addresses flushing happening and a flood of messages being sent to each member of our bond. Unfortunately due to time constaints and other things we cant take the linux kernel that discards messages on the standby member of the bond. So we set out to fix this within our network. We instituted a ping to a pingable gateway on each server to keep mac addresses up to date. So finally we got hit by rapid spanning tree. There was no way we could ping fast enough to beat rapid spanning tree.

So we decided to move the problem within our switches. We increased the cost of the uplinks to 100 on each of the 66 and 67 vlan access ports on switch1B to the lab switch. This causes us now to have switch1A access ports active, and the gigabit crossover between switch1A and switch1B active but the uplinks to switchB get blocked now because their cost+priority is higher now.

This has cleared everything up. No dups. No problems. One anomaly we see though is the mac addresses for the 67 vlan for the linux boxes will not stay in the lab switch above switch1A and switch1B. Switch1A and switch1B mac address caches remain up to date always with the pingable gateway we are running every 10 seconds. But in the lab switch they are connected to I can do a show cam dynamic searching for the mac and it might show me the mac once out of every 10 times. So I know its getting in there then being flushed.

I've turned on "set logging level spantree 7" but i see nothing in the show logging buffer that would lead me to believe any tcns are happening?

Anybody got any clues?

wong34539 · ‎07-17-2006

The Add option allows you to add Ethernet MAC addresses for devices that might pass traffic through the bridge. If no addresses are added through the Add option, the bridge learns the first eight MAC addresses that pass through its Ethernet Port. Subsequently, only data from those addresses is allowed to pass through the bridge.

Caution: The first MAC address you add should be that of the PC you are using to Telnet or browse to

the bridge.

You should add MAC addresses if there are more than eight Ethernet devices attached to the hub to

which the bridge is connected. This ensures that the selected devices communicate through the bridge. After an address is added, the bridge won't learn any more addresses. You must type each MAC address you wish to have communicate through the bridge (up to eight).

Once you enter the first MAC address, the MAC addresses of every other device that you want the

bridge to communicate with must be entered. The process is not automatic and the bridge will no longer "learn" any addresses. The addresses must be manually entered.