best practise for switches replacing order?

[baselzind]

I have cores 6500 , server farm 4500 and edge switches 3xxx/2xxx. I will be replacing them with cores 9500 , server farm 9400 and edge switches 9200. is there a best practice for specific order on which type of switches to replace first , cores , server farm or edge switches? 

[balaji.bandi]

Its all downtime you getting and Rack space available for you.

If you have enough space available in the RACK.

1. i will build the new setup in line - CORE and Server farm switch / access switch

2. Connect Core to Core Layer 2 / Layer 3 depends on your requirement

3. Make some testing all working Gateway Routing on the new setup Core and Server farm and Access switch.

4. Pick up least effected area for Access swtich area and Migrate to new Switches and Monitor - Increase this Migration gradually until all Finsihed - same case with Server Farm - moving the Servers and test it.

Once all Moved then you need small downtime depends on the setup, Move Core Uplink and any SVI from OLD Core to new Core (one at a time, of big bang )

Monitor is key point here all the time and make sure you take feedback from Service desk and your NMS monitor and Logs of new device see make sure it working as expected.

This rough you can go deep each area how you looking to test.

[baselzind]

my issue is that i don't have rack space , any idea how the process will be if so?

[balaji.bandi]

If no space - there is many thinkgs need to consider

1. If you Upgrade any part of the device, is that compatable and works as expected (this need to be tested)

2. You can choose which ever model Top down or down top (like CORE - Server --Access switch) or Access switch - Server farm - CORE - the risk level same.

3. You need to Look for the uplink ports and configuration compatability with IOS XE

4. I suggest to make a POC lab and test it by connecting to exiting CORE to CORE see if that works.

[baselzind]

what will i gain from connecting the old core to the new core?

[ammahend]

if you have rack space then follow top down approach, extend an L2 from current core to new core, move all IDFs to new core, at the end move SVIs and routing to new core, what you gain is flexibility to take your time and do one location at a time and production effects are minimum.

if you don't have rack-space, follow down to up approach where you upgrade IDFs first and do core at the end overnight or on weekend.

as suggested check if your new access layer can work with existing core, a small POC would be a good idea in this case.