02-06-2019 12:07 PM
Migrating a VM farm from an old network (subnet A for simplicity) ... this farm connects to a Cisco 6500 switch on its own vlan / subnet.
I will add a new Aruba 8320 switch to this existing infrastructure in readiness to eventually take over the role of the old Cisco 6500 switch (which holds the SVI - default gateway for the VM farm). I’m just a little stuck on this VM farm migration ... I need to migrate it to new hardware with as little down time as possible ... use the same address space when it transits over to the new switch / hardware.
I’m thinking of making a trunk link between the 8320 and Cisco 6500 ... add same vlan / SVI on 8320 and just quickly remove the SVI on the old Cisco switch ... so effectively move the VM farms default gateway to a new switch ... then migrate the farm across to new hardware / storage.
Bit of a ramble but is this the best way - feels dodgy to me but I can’t think of a better way?
Solved! Go to Solution.
02-28-2019 08:32 AM - edited 03-04-2019 03:01 AM
I tested such scenario. It is even smoother:
- do the L2 trunking connection as you described.
- set-up the same IP than on old 6500 for the VSX active-gateway VIP on each L3 SVI.
- when you issue active-gateway command, a Gratuitous ARP will be sent. But it is likely than one is not enough.
- you may have still part of your VMs having 6500 MAC as default-gw. Start GARP on 8320. As this is not in CLI, here is how:
8400:~$ sudo su
8400:/home/admin# ip netns
ntb (id: 0)
Identify the proper VRF (VRF_1 is mgmt, and then in the order of creation).
2 proposed ways:
a) ip netns exec VRF_3 ifconfig
and check the IP addresses confirming the VRF ID.
b) use VRF API and look for "table_id" and associated "name".
c) switch:/home/admin# ovsdb-client dump VRF name table_id
Once VRF ID is identified, start bash:
8400:/home/admin# ip netns exec VRF_3 bash
ifconfig to identify SVI name (here below I want to send gratuitous ARP from 10.1.10.1
which is the anycast IP address of the VSX active-gateway)
vsx03f21 Link encap:Ethernet HWaddr 00:00:00:10:10:01
inet addr:10.1.10.1 Bcast:0.0.0.0 Mask:255.255.255.255
Source ARPing from the said SVI as a L3 broadcast:
8400:/home/admin# arping -I vsx03f21 -s 10.1.10.1 -bA 10.1.10.255
You may tune this command:
- replace broadcast destination with single host IP (ex: 10.1.10.101)
It works like a charm to refresh hosts/endpoints ARP cache with the VMAC of the active-gateway
of your new VSX cluster, replacing old gateway. Check ARP on migrated hosts.
Shut L3 SVI on old switch (assuming new CX takes care of upstream routing).
Stop ARPing on this subnet.
Repeat the process per subnet.
Unicast me if not clear.
This process will be documented in future.