Stacked switch vlan interface (SVIs) routing and load balancing - ip cef ( can it really works ) ?

[SJ K]

Hi all,

There are 2 coreswitches in siteA and they are stacked together

There are 2 coreswitches in siteB and they are stacked together.

The coreswitches will be connected in such way

SiteA CoreSwitch 1 <-----> SiteB Coreswitch 1

SiteA CoreSwitch 2 <-----> SiteB CoreSwitch 2

Please see below network diagram ->

Right now, the coreswitch are represented as just 1 per site (they will be stacked on actual site). I have enable eigrp for all the connected network and PC can ping to 1 another without any problem.

However, there is 1 particular portion that i would like to seek advices from gurus here (please look at the portion on the connectivity between the core switches -- 192.168.254.0/24 network)

I have not used any routed port, but plain vlan interfaces for the connectivity between the 2 core switch. In this setup, if there is another normal switchport in vlan4 on coreswitch1, can it reach another switchport in vlan4 on coreswitch 2 as though they are in the same vlan ?

-- though everything is working fine, the confusing portion here that it seems like i have 2 SVIs on the same vlan, i am not sure if packets in the same vlan on such setup above will still be switched ?

q1) Is my setup above sound/correct ?

q2) like i have mentioned earlier on, my coreswitches are stacked on each site. (Imagine now the coreswitchs in my diagram above are stacked on each site)  and  i will be using a vlan interface (svi)  across the 2 switches in a stack to do routing.

The reason i use a vlan interface/instead of routed ports is that

1. i can have just 1 svi interface across the 2 switches in the stack, instead of 2 routed ports on each switch in the stack.
2. if 1 switch in the stack is down, my svi interface still works and traffic still can be routed across sites.

The ultimate problem is that though  I have 1 SVI across the 2 switches in a stack, there are actually 2 physical wires from SiteA stackswitch to Site B stackswitch.

In show ip route on Site B stackswitch

D    192.168.1.0/24 [90/51225600] via 192.168.254.2, 02:37:03, Vlan4

You can see that the exit interface is Vlan4.

How do i know which "wire"/physical interface/switch in the stack does vlan4 actually use to send the traffic ?

I have ip cef turn on, can load balancing happen across the "Vlan4"  interface ?

Regards,

Noob

[Carlos Villagran]

Hi!

They will be able to communicate, yes since I want to assume the link between the core switches is Trunking so yes, no problem with that, hosts in the new VLAN 4 will communicate even if they exist in the switch core 2.

1. It is correct although it is not a best practice to have VLANs traversing to different core areas. 

2. If you are connecting two cables between your core switches and this is NOT a routed interface, then be aware that one of the interfaces will be blocked by STP. You can use the show spanning-tree vlan x command in order to know which cable is being blocked. The one that is NOT blocked is the one actually passing traffic, now, I would suggest an etherchannel configuration instead so the bandwidth is used efficiently. 

Hope it helps, best regards!

JC

[SJ K]

Hi Carlos,

Thank you and greatly appreciate your advice on the topic !

Back to the questions ->

They will be able to communicate, yes since I want to assume the link between the core switches is Trunking so yes, no problem with that, hosts in the new VLAN 4 will communicate even if they exist in the switch core 2.

q1) the connection between the 2 core-switches are actually not trunk, i even tried setting them up with different VLAN number (coreswitch1 vlan10, coreswitch2 vlan4) and it still works ( for the routing portion ). But eventually i set them up as both VLAN4;  and hence my doubt on - if hosts added in on those VLANs will still be able to communicate(switched) as though if they are in the same VLAN  (if the point to point connection are actually in different VLANs).

2. If you are connecting two cables between your core switches and this is NOT a routed interface, then be aware that one of the interfaces will be blocked by STP. You can use the show spanning-tree vlan x command in order to know which cable is being blocked. The one that is NOT blocked is the one actually passing traffic, now, I would suggest an etherchannel configuration instead so the bandwidth is used efficiently. 

q2) you're so right ! - i totally forget about the spanning-tree.  If i were to configure the two cables between the coreswitches as etherchannel ->

1. should i configure them as a L3 port-channel or should i still stick to using a vlan interface ? or either both will do ?
   
   
2. there is only a 1:1 server communication between the 2 sites (production db to standbydb); thus always the same source and destination - , but i need both links to be utilized.
   With ip cef -> can i check if the next hop / exit interface is still a single interface (vlan/port channel) in the coreswitch cef table ? and i am not able to achieve/issue load-sharing per packet on the vlan/port-channel interface ?

In short, can i use per packet load sharing using ip-cef  on the (vlan or portchannel interface) on the above setup ?

if not; how do i load share per packet via the 2 links between the core-switches in SiteA and B ?

Really eager looking forward to your thoughts!

Regards,
Noob

[Carlos Villagran]

Hi Koh!

a. You can use either a L2 or L3 port-channel, this depends on your design and needs. For example if you want the servers to be in the same broadcast domain/subnet then a L2 etherchannel would be needed, the L2 will give you the use of both links however be aware that a VLANs traversing core switches is not a best practice. 

If you use a L3 only one link will be used and the other one would be used as backup/redundancy.  So it seems that a L2 etherchannel would be a better election if you really need to use both links.

b. There is a procedure to calculate which link is to be used in the etherchannel bundle. There are commands too which tell you exactly which link ia taking but those are platform dependent.

Please refer to the following links for the load-balancing algorithm calculations:

https://supportforums.cisco.com/blog/150511

Hope it helps, best regards!

JC

[SJ K]

Hi JC,

Thanks for your wonderful reply.

q1) Can I check does L2 etherchannel /port-channel forward broadcast ?

If you use a L3 only one link will be used and the other one would be used as backup/redundancy.  So it seems that a L2 etherchannel would be a better election if you really need to use both links.

q2) Why would a L2 port-channel use both links while a L3 port-channel will only use one link ?   I thought the usage of the links are tie to the load-balancing algorithm.

Since there is only 1:1 src and destination (load balancing via src-dst mac (L2) or src-dst ip (L3)) would just use 1 link right ?

Can you elaborate further ?

Regards,
Noob

[SJ K]

Hi Carlos,

Still there ? ;(

Regards,

Noob

[Carlos Villagran]

Sure, please allow me some time to write my answer. I have been having some difficult days at job. 

Regards!

JC

[SJ K]

Hi Carlos,

I am sorry to hear that.. 

I will wait here for your reply when you are less loaded.

Thanks!

Regards,

Noob