Migrating from HP Procurve to Cisco

[zekebashi]

Hello, 

See attached.  We're planning on migrating our client from HP Procurve switches to Cisco switches/routers. Here's the current and new topologies: 

Current topology: 

- Headquarters:  

- hp Procurve Switch(hp-Core): This is the Core switch which links to 13 remote locations via point-to-point L3

- Static routing is in use

- There 10 servers(ESX, Vmware, or others) that are uplinked to this Core switch 

New topology: 

- Headquarters: 

        - Cisco C4500-X switch(Cisco-Core): This will be the new Core switch and will replace the role of the hp Procurve, so all 13 remote locations will be linked to this Core switch.

        - OSPF routing is configured(this sw is the backbone Area 0)

       -  Cisco C4500-X switch(Cisco-Dist): This will be the distribution switch uplinked to the Core switch via a Trunk. All 10 servers will be uplinked to this switch. OSPF is enabled on this sw        this is in area 1

A consulting engineer recommended the following migration plan and I wanted to see someone has had an en experience migrating from HP to Cisco and if they see any issues with this plan or have any other recommendations. 

Recommended migration plan: 

- Since the hp-Core doesn't have OSPF enabled, the Cisco-Core will be linked to hp Core via L3, enable OSPF on hp-Core, and add the default information originate always statement,

- Uplink the new Cisco-Dist to the hp-Core via L2(Trunk), and then start to moving the servers to the new Cisoc-Dist switch.

Much appreciated. 

Best, ~zK 

[Masoud Pourshabanian]

You are most welcome. Let us know if you have more questions.

Masoud

[zekebashi]

In terms of routing it sounds like you are not running a dynamic routing protocol between the HP and remote sites. ----> That's correct. Currently, there's no dynamic routing running on the HP core and all the HP remote switches. 

If this is the case then do the remote sites just use a static default route pointing to HP ie. no local internet at sites. ----> Not really clear here! So, should we enable OSPF on HP Core and all remote sites or just the HP Core, leave all static routes on the HP Core, and add static route pointing to HP Core?

And then presumably the HP has static routes for each remote site ? ----> That's correct. The HP core has static routes to each remote site. 

If so when you migrate a link you can move the HP IP to the new core switch but then you have to add routes to the new core switch for the remote sites subnets and make sure they are redistributed into OSPF so the HP switch knows about them. --> I'm not sure I am clear on this point here. Here's the steps I plan to take: 

     -  Cisco Core has OSPF area 0 configured

     - Cisco Core has 13 L3 interfaces(admin shut) configured with the same IP addresses as the ones currently being used by HP Core and remote sites(ie: 10.11.1.1/24 & 10.11.1.2/24)

      - Cisco Core: OSPF networks advertised are the point-to-point links(ie; 10.11.1.0 0.0.0.0 area 0)   

      - Link the Cisco Core to the HP Core via L3 

     -  Enable OSPF on HP Core 

     -  Add default information originate always statement to HP Core 

Do I need to enable OSPF on the remote switches? 

Once you have migrated to the new 4500-X core you the would not need to redistribute into OSPF because your core switch will be sending a default route to the distribution switch I assume. 

[Jon Marshall]

The remote sites must know how to get to the central site so if you are not running a dynamic routing protocol they must be using static routes, probably a default route.

If it is working now you don't need OSPF on remote site switches if you don't want to.

The point I was making is that if a remote site is connected to the HP then your new switches can get to the remote site because the HP switch is sending a default route to the new switches.

If you move a remote site to the new switch you would then remove the static routes for that site off the HP and add to the new switch which means the new switch can now reach the remote site.

But if traffic from a remote site comes into the new switch and goes to the HP switch or beyond to the firewall when the return traffic gets back to the HP switch it won't know where to send the traffic.

So on the new switch when you add the static routes for the site you need to redistribute into OSPF so the HP switch knows to send the return traffic to the new switch.

It really depends on what routes are at the remote sites.

Jon

[paul driver]

Hello All

I have been following this post with some interest and working on Jons/Massouds suggestions and that of your consultant- I went a head an Labbed this one up last night?

Goal:
Migrate remote sites onto CIsco-Core and ESX hosts onto Cisco-DIstribution
remove HP core



Pre-Migration

HP Cores:
L3 to Fw ( WAN) via default route
L3 to remote sites ( no igp)
SVI for ESX Host

Remote sites
L3 to HP Core ( no igp just static default towards HP Core)


Intial setup between Hp and Cisco

1)
Created L3 between HP & Cisco CORE
enable ospf between them (area 0) and advertsie default from HP Core

2)
Created L3 between Cisco CORE  & Cisco DIst
enable ospf between them in area 1
making Area 1 tottaly stub area -  ( this advertsied a default into area1 stub area for the Cisco-Disbtrubtion access)


3)Create L2 between CIsco CORE and HP CORE ( this is for ESX migration -  - one SVI  at a time)


The intial setup worked accordinlly , remotes and esx host connectivity working fine and no interutption/outage occured.

Migration - ESX Hosts
I moved  one host at a time across onto Cisco distribution, keeping the L3 for this vlan on the HP, and Due to l2 connection exisitng between Cisco distribution core and HP core connectivity remained

Once all hosts were moved, I relocated the L3 SVI onto the Cisco distribution core , disbabling it on the HP core and advertised its subnet into the ospf process,

repeated process to all ESX SVi vlans

Migration - L3 remotes sites
I relocated each remote site physcal L3 connection onto the Cisco-Core,
added L3 addressing to the new routed port and advertsied it subnet into its ospf process within area 0



Once all remote sites and Esx hosts had been relocated, I then procceded to migrate the routing and WAN connection over to the Cisco-Core

1) Shutdown the L2 between  CIsco CORE and HP CORE
2) Shutdown the L3 between  HP & Cisco CORE
3) Relocated the WAN connection over to Cisco-CORE change static default route I had pointing at WAN FW

Obvioulsy the outages incurred were the relocation on the L3 addressings at time of cut over

So to summerise Jons and massouds suggstion worked occordingy.

hope this was of help

res
Paul

[zekebashi]

Outstanding. This is great work, Paul! 

Thanks so much. This is really helpful. I really do appreciate your help with this. 

Best, ~zK 

[Masoud Pourshabanian]

zekebashi

servers will be moved to the distribution switch once at the time and corresponding interface VLAN will be removed on HP and will be created on distributed switch. It is the reason for the trunk interface; however, it can be converted to l3 interface after migration.

Masoud

[Masoud Pourshabanian]

Hello,

Te recommendation sounds good,  but there is no indication how you want to remove HP switch. You need to consider it in your plan. It depends if you are activating OSPF on the device located above the HP switch or not.

Masoud

[zekebashi]

Thanks, Masoud. 

The proposed recommendation was that after we've ensured that routing between all VLANs and all networks have been verified, then we will simply start moving the links of those 13 remote sties to the new Cisco-Core sw.  Any thoughts?  

The device above the hp-Core is a FW, so, no, OSPF won't be activated about the hp-Core. 

Best, ~zK 

[Masoud Pourshabanian]

I do not see any problem in migration of 13 remote sites. I suppose 3 switches are close to each other so physical connectivity will not be a problem. Before migration, test portchannel between firewall and cisco switch to make sure you can remove HP fast. Take a look at Jon suggestion also to get more idea.

Masoud