Migrating Virtual Servers to another LTM pair with newer version is slow

[Ronens]

Hi there, like the title describes, I've encountered a problem when I migrate VS's from an LTM-VE pair that are in version 11.1.3 to another pair of LTM-VE that are booted with version 14.1.2.8.

The way I'm doing the migration is as follows:

First I create exactly the same nodes, node pools, monitor and the VS as configured in the old pair but the VS is with a different ip.

After I check that the VS is responsive on the new LTM, I change the ip of the VS on the older pair, then change it in the new pair to what it originally was before in the old pair.

The problem that occurs is that after changing the VS ip address on the new pair to what it was on the old pair, it takes about 6 to 10 minutes until the VS is responsive.

If I change the ip addresses back to what they were before, the VS comes to life on the old pair instantly.

Also when I'm waiting for the VS to come up on the new LTM pair I can see that the original VS ip still exists on one of the old LTMs, even though the GUI shows the temporary ip I configured for the VS.

What can I check in this situation?

And is it possible this happens because the versions don't match?

Thanks in advance.

[Samir]

I am suspecting it is not because of F5 LTM, May be other factors you need to check. It will give more clue to troubleshoot and correct the configurations.

Example

1. Trace-out both IP(Applications), I mean old Setup and new Setup.
2. Check the firewall/Router if you need any additional policy.
3. if you have windows/Unix machine try to run the command #time curl -Ik https://<Applicationname>
4. Customise the HTTP & TCP profile in v14.x setup(Basically it will help to make application faster)

Hope it will help to drill down the issue.

[Daniel_Wolf]

My first guess would be looking at the ARP table of the connected switch.

Sounds like the IP address of the VS is still in the ARP table with the MAC address of the old BIG-IP pair.

If this is your problem, also take a look at K13502: Configuring MAC masquerade (11.x - 15.x).

[jaikumar_f5]

I agree to what Daniel has stated, It should be because of the ARP entries not updated on the upstream device.

As you say you are migrating, one possibility I can think of is to disable the ARP on the respective VS address during your migration and see if the ARP gets updated from the new device automatically. Also disable the VS during migration. Keep us posted.