Catalyst 2900 and 2950 switches and multicast

[zephyran]

Our network is pure NT 4.0 with a backbone of 3 Catalyst 2900, and one 2950, switches (all 24-port 100BaseTX Ethernet). We also have a PIX Classic running the version 4.0.6 IOS.

We use Symantec Ghost to re-image our machines. Ghost uses a multicast that is accessed by machines that have been booted using a special network floppy.

Whenever we run the Ghost multicast to re-image one or more machines over the network, the Internet gets blocked out from all machines (with "Page cannot be found" errors), from the second the multicast starts till the second it stops. This started happening primarily after we added the 2950 switch to the 3 2900s.

I've looked into issues with the firewall and multicasts, but haven't found much of anything helpful, and no one in the PIX forum can help. Could the switches somehow be part of the problem, and is there a way to check/fix it?

By the way, the switches are minimally-configured, because of a shortage of internal IP addresses. They are configured with IP addresses that are outside the subnet used by the LAN, and are not configured in a cluster. Could this be part of the problem?

 

[munsonbrown]

Here's something to try. Let's say you have 3 switches (1,2, and 3). 2 and 3 both connect back to 1. The PIX and the Ghost server are also in one. The users receiving images are all on switches 2 and 3. Can a user in switch 1 browse the web? Ghost has a tendancy to use all available bandwidth. If the transfer bogs down the uplink, other users behind the uplink will suffer. Would it be unacceptable to set the Ghost server port for 10 meg?

Some things to think about. Good luck.

 

[tduplantis]

I have had basically the same problem on my network, except I have 8 2924XL switches, and when we run Ghost, 3 ports on my server room switch start blinking orange and the network slows to a crawl. The ghost server is on a switch downstairs, the server room switch is upstairs. I had to take the ghost server off the network and set it up on its own private lan. If you have resolved this, please let me know what you did.!!!

 

[zephyran]

What I've ended up doing was what munsonbrown suggested: I set the speed for the server's link (in the switch) at 10 MBps, full-duplex. Not the best solution, but it seems to have cut out the problem.

 

[jgercken]

Just a guess here. A few switches understand IGMP but most need a router to give them CGMP messages indicating which MAC addresses have subscribed to multicast channels. It sounds as if this isn't configured properly and the switches, not knowing any better, are simply flooding all the multicast packets out all ports. The effect is your network is choked.

-Jeff

----------------------------------------
Wassabi Pop Tarts! Write Kellogs today!

 

[zephyran]

Well, it's all happening on a LAN with the Catalyst switches as a backbone. Do the switches need to be configured to fisx the problem?

 

[teninch]

My suggestion to you if you are having insufficeint IP problems is to install and configure a DHCP server. Exempt the first 25-30 ip adressess from the DHCP pool. Utilize those addresses to statically configure your switches, printers and other servers on your LAN. Also try not to have your server (PDC) connected directly to your master switch if you have 2912 MF -XL or other Cisco 29x FDD distributing down from a Fiber backbone.

 

[zephyran]

I'm not having insufficient IP problems...I think you posted to the wrong thread