Intergrating HP Blade servers into a Cat 6000 backbone PART2

[michael4702]

Please reference an earlier post of mine "Intergrating HP Blade servers into a Cat 6000 backbone"

From further research I have come up with some additional questions. I need to be water tight on this as I could take down a 2000 user network if im wrong.

First and probably the easiest, how do I find out which of my switches is the root bridge?

With the above question in mind, the core of my network is 2 cat 6000's, which are linked by a 3 gig ether channel. This either channel is the backbone of our network. Now logic denotes that one of the two 6000's is the route bridge which from the above question I hopefully will soon find out. But could both 6000's be the route bridge? The designers of our network have long since left but we do seem to have a procedure in place where vlan's with even and odd numbers are set up differently... So could the 6000's be doing half and half\load sharing?

Next question, assuming at least one of the 6000's is the route bridge and therefore the centre of my spanning tree network, will simple insuring that the bridge priority on that switch is higher than anything else stop the blades from becoming the backbone. The reason I ask this question is because I have discovered port costs. And the port costs for going from one 6000 core switch across the backbone of the blade server to the second 6000 core switch may be better than going across the 3 gig etherchannel between the two core 6000’s. If the above paragraph has made sense! The question is, do I need to assign higher port costs to the ports\ether channels that connect to the blades therefore insuring the ports\ether channels that currently directly connect our core 6000’s continue to be the backbone?

Regards and thanks.

Michael.

[Roberto Salazar]

You said: First and probably the easiest, how do I find out which of my switches is the root bridge?

****

if CatOS: show spantree summary

if Native IOS: show spanning-tree summary

****

You asked: With the above question in mind, the core of my network is 2 cat 6000's, which are linked by a 3 gig ether channel. This either channel is the backbone of our network. Now logic denotes that one of the two 6000's is the route bridge which from the above question I hopefully will soon find out. But could both 6000's be the route bridge? The designers of our network have long since left but we do seem to have a procedure in place where vlan's with even and odd numbers are set up differently... So could the 6000's be doing half and half\load sharing?

****

Catalyst uses Per Vlan STP, so yes, it is possible that one switch is the ROOT bridge for odd vlans while the other is the ROOT for EVEN vlans.

You added: Next question, assuming at least one of the 6000's is the route bridge and therefore the centre of my spanning tree network, will simple insuring that the bridge priority on that switch is higher than anything else stop the blades from becoming the backbone.

****

You mean LOWER the priority. The lower the priority the better chance the switch will be the ROOT for that vlan or those vlans.

You're further inquiry:The reason I ask this question is because I have discovered port costs. And the port costs for going from one 6000 core switch across the backbone of the blade server to the second 6000 core switch may be better than going across the 3 gig etherchannel between the two core 6000’s. If the above paragraph has made sense! The question is, do I need to assign higher port costs to the ports\ether channels that connect to the blades therefore insuring the ports\ether channels that currently directly connect our core 6000’s continue to be the backbone?

****

The STP will calculate the path cost for you. You should not need worry about it as long as the switches agree who the root is. However, it is good to understand the cost to confirm optimal path selection.

****

[michael4702]

Thankyou for your response.

The last problem/query i have is from the last paragraph of my last post, please see below.

You're further inquiry:The reason I ask this question is because I have discovered port costs. And the port costs for going from one 6000 core switch across the backbone of the blade server to the second 6000 core switch may be better than going across the 3 gig etherchannel between the two core 6000’s. If the above paragraph has made sense! The question is, do I need to assign higher port costs to the ports\ether channels that connect to the blades therefore insuring the ports\ether channels that currently directly connect our core 6000’s continue to be the backbone?

****

The STP will calculate the path cost for you. You should not need worry about it as long as the switches agree who the root is. However, it is good to understand the cost to confirm optimal path selection.

This is the problem, if STP calculates the path costs for me and determines that the blade server route is best, then it will use the blade server route as the back bone. I do not want this. I want to insure that the direct link between the core cats continues to be the backbone... Even though the blade route may be the fastest I do not want the entire network using this route as it will greatly reduce the bandwith avalible to the servers in the blade.

Can I use port priorities to over ride STP? Therefore making sure STP see’s my direct link between the core cats as the best route?

Many thanks.