Work Load Migration from Non NSX to NSX Platform - Down time Consideration

[rajeevsrikant]

I have my scenario where i need to migrate work load from non NSX platform to NSX platform.

The option is to use L2 Bridging & migrate the work load from non NSX to NSX platform.

The question is related to the down time.

Till I change the default gateway(D.G) from Physical L3 switch to the NSX DLR there will be no downtime.

When the D.G is changed from Physical L3 switch to the NSX DLR what is the normal down time it will take.

What are the considerations that should be taken into account when the switch over is done, in terms of MAC Address, ARP entry etc.

What is the best way to go about it in terms of minimum down time ?

[hansroeder]

I'm sure some people who have experience with these migrations will post something about their personal experience, so that might point you in a semi-right direction

However, since every environment is different (different routing protocols, different hardware with different timeouts, etc.), my advise would be to test this for yourself. I would create a dummy subnet, put the default gateway for this subnet on the L3 switch, and then do a test migration. This, in my opinion, will give you the best possible answer for your own environment. This will also give you the opportunity to improve on your migration design, as you are bound to run into some problems (it's still networking right?! ).

I'm not quite sure what considerations should be taken into account in terms of MAC, ARP, etc. This should all sort itself out. The default gateway for the VMs in the subnet that you're migrating will change to the DLR LIF, so that also means that it will get a new MAC address. But this is not something you need to configure. Regarding L2 communications between the VMs in the subnet that you're migrating, nothing really changes. So it's really the L3 stuff that you need to thoroughly work out.

[rajeevsrikant]

Thanks. I understand that it depends on individual environment & it needs to be tested.

Basically wanted to know how others have done it & what impact or downtime they encountered  in their environment.

Also was interested in knowing what specific actions or changes they did to reduce the down time.

Regarding the question on MAC & ARP would like to know is it required to clear the MAC & ARP entries in the physical switches or in the servers to have faster convergence to the new D.G which will be the DLR.