11-13-2020 12:18 AM - edited 11-13-2020 12:20 AM
HI all,
hope to find everyone well
Recently was put up with this challenge at my work place.
I have a campus network with around 25 Ciscos 3560 with the IPBase license on them. Every switch is handling the same vlans (10,20,30,40) but different networks between them. One has the network 10.8.20.0, the other that is in another part of the network has 10.9.40.0, another 10.14.40.0, and so on.
Now I was asked to create the following, I have a machine on the network 10.14.20.0, and I'm going to have a backup server running on the network 10.9.20.0 that will be making Hyperv Replicas of the Virtual Machines running on the machine located on the 10.14.20.0. Now due to the nature of Hyperv and the system, if the primary machine fails the backups quicks in and will start exactly with the same IP address of the machine located on the network 10.14.20.0 in order for all the systems that rely on this IP to continue working as they should.
Now my issue is, how can I take this VLAN 20, from one switch to another switch that already has a VLAN 20 on them and with a totally different network?
Is there any way of doing this, or do I need to create a totally new vlan and expand this vlan in order to have connectivity across the network and for the IP addresses to remain the same?
Hope to have explained the issue well
Thank you before hand for the support
Solved! Go to Solution.
11-13-2020 12:32 PM
Thanks for the additional information. So as I understand it there are to be multiple IP addresses that normally will be present on switchA but in case of a failure of the server the addresses will suddenly appear on switchB which is in a different part of the network. As long as there are not switch to switch connections on access ports in vlan 20 and not switch to switch connections on trunk ports carrying vlan 20 then it really does not matter that both switchA and switchB both have a vlan named 20. What does matter is that you want some addresses in subnet 10.14.20.0 to appear on switchA but also sometimes to appear on switchB. And that is a challenge.
GRE is a common way to establish a link to connect two locations. But GRE assumes that there is one subnet at the A side and a different subnet at the B side of the tunnel. It does not do well when you want the same IP addresses to appear at A and also at B. (and I am not sure that they are support on your switches any way.
L2TPv3 would be a better way to extend the vlan but I do not believe that it is supported on your switches.
One way of looking at the issue is that you have one broadcast domain for vlan 20 on switchA and another broadcast domain for vlan 20 on switchB. If you have some IP addresses that usually will be on switchA but sometimes will be on switchB then to make it work both switches will have to participate in a single broadcast domain. One way to achieve that would be to configure trunk connection carrying vlan 20 between switchA and switchB that would carry vlan 20. But that trunk would run through several other switches and would impact those switches. It seems to me that the better solution would be to configure a new vlan on both switchA and switchB, configure trunk connection between the switches to carry the new vlan, configure a new subnet on the new vlan (same subnet on both switches for that vlan) and to put the server and the backup into the new vlan/new subnet. That way the IP addresses can move between switchA and switchB without any problem.
11-13-2020 02:52 AM
- Basically this can't be done. The idea is that a vlan spans over network components but remains on a unique subnet addressing scheme, so that pure layer2 communication remains possible for all hosts participating on that vlan.
M.
11-13-2020 03:38 AM
Thank you for the reply Marce
So there's no way of creating any type of tunnel?
11-13-2020 08:30 AM
I am a bit confused about this topology. I think it is saying that vlan 20 exists on some switch with a particular subnet and that another switch in the network also has vlan 20 but the IP subnet on the second switch is different from the subnet on the first switch. I assume that at this point the 2 instances of vlan 20 are separated and now the original poster needs to connect them. If I have not understood correctly then please provide clarification.
It seems to me that the 2 instances of vlan 20 can be connected and the vlan interface(s) for vlan 20 could have an IP address command for the first subnet and could configure a secondary address on the vlan interface(s) for the second subnet.
11-13-2020 09:18 AM - edited 11-13-2020 09:19 AM
Hi Rick,
So basically the switches between themselves are connected with another vlan, let's say vlan 180 to connected between two switches, and so on, and inside the switches they have community vlans, vlan 20, vlan 30, 40 and so on.
Basically all the traffic stays located on those vlan and then it gets routed if it needs to go another network, basically we can have a computer, pull the traffic of a device 10.9.20.1 and its routed 5/6 or mote times between the switches until it reaches the respective computer.
Sorry I should have mentioned this
11-13-2020 10:03 AM
I am not clear about some things. You say you have a machine on the network 10.14.20.0. Do you need the entire subnet to appear on the other switch? Or do you need the individual host IP to appear on the other switch?
Also you mention things being routed across the network. What is doing this routing? Is it using static routes or is there some dynamic routing protocol? I am wondering if there might be some way to dynamically advertise a /32 host route when the backup copy begins to run.
11-13-2020 10:47 AM
All the routing is being done via the 3650s using EIGRP
I have a server running three VMs on the subnet 10.14.20.0, this server plus the VMs occupy 4 addresses. Then I'm going to have another server (backup) running VM replicas, on the subnet 10.9.20.0, because this VMs are replicas in case the main server fails, this backup VMs will boot up with the exactly same addresses because they are a mirror of the VMs of the primary server on the subnet 10.14.20.0. I needed to put in some way two ports available for the network 10.14.20.0 that is located in a totally different switch of the network. Basically a tunnel of some kind... Dunno if this is possible...
I have some colleagues that talked about GRE and L2TPv3, but for what I saw think the 3650s don't support it and to be honest I don't even know how to apply it.
11-13-2020 12:32 PM
Thanks for the additional information. So as I understand it there are to be multiple IP addresses that normally will be present on switchA but in case of a failure of the server the addresses will suddenly appear on switchB which is in a different part of the network. As long as there are not switch to switch connections on access ports in vlan 20 and not switch to switch connections on trunk ports carrying vlan 20 then it really does not matter that both switchA and switchB both have a vlan named 20. What does matter is that you want some addresses in subnet 10.14.20.0 to appear on switchA but also sometimes to appear on switchB. And that is a challenge.
GRE is a common way to establish a link to connect two locations. But GRE assumes that there is one subnet at the A side and a different subnet at the B side of the tunnel. It does not do well when you want the same IP addresses to appear at A and also at B. (and I am not sure that they are support on your switches any way.
L2TPv3 would be a better way to extend the vlan but I do not believe that it is supported on your switches.
One way of looking at the issue is that you have one broadcast domain for vlan 20 on switchA and another broadcast domain for vlan 20 on switchB. If you have some IP addresses that usually will be on switchA but sometimes will be on switchB then to make it work both switches will have to participate in a single broadcast domain. One way to achieve that would be to configure trunk connection carrying vlan 20 between switchA and switchB that would carry vlan 20. But that trunk would run through several other switches and would impact those switches. It seems to me that the better solution would be to configure a new vlan on both switchA and switchB, configure trunk connection between the switches to carry the new vlan, configure a new subnet on the new vlan (same subnet on both switches for that vlan) and to put the server and the backup into the new vlan/new subnet. That way the IP addresses can move between switchA and switchB without any problem.
11-13-2020 02:20 PM
This was an interesting and challenging situation. I am glad that my suggestions have been helpful. Thank you for marking this question as solved. This will help other participants in the community to identify discussions which have helpful information. This community is an excellent place to ask questions and to learn about networking. I hope to see you continue to be active in the community.
11-13-2020 02:22 PM
Thank you very much Rick for your extremely insightful explanation!
I will do that then, the costumer will not be happy but it's the only thing that can be done.
I will have to program an entire side of the network (10 switches) with this vlan and a new subnet in order for everything to communicate and to create several redundant paths. Also I will need to program the Spanning Tree, and to program HSRP for this vlan... Have a little bit of work ahead of me...
Thank you very much once again!
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide