Add known issue for IP address change upgrading to v1.7 when using DHCP

[tserong]

Problem:

If you are using DHCP to configure your host IP addresses, it's possible the IP addresses may change during upgrade to v1.7.0, which will prevent the cluster from starting correctly. This requires manual intervention to remedy.

Solution:

The IP address change can be fixed by editing the generated NetworkManager connection profile to include the old (pre-upgrade) DHCP client ID. Note that the example files included in the docs here are deliberately shortened (lines removed indicated by ...) to try to highlight only the relevant parts.

Related Issue(s):

harvester/harvester#9260

Test plan:

See reproducer in harvester/harvester#9260. Try applying the workaround as documented here in that environment. It should work :-)

[github-actions]

Name	Link
🔨 Latest commit	dacf172
😎 Deploy Preview	https://691fd2f5d4ae945dc086f0eb--harvester-preview.netlify.app

[ibrokethecloud]

The actual doc looks good. Just wondering if we should advise users to check this before upgrade and write the 91-NetworkManager file before triggering the upgrade? If a user runs into issue during upgrade and node is only accessible via a serial console, generating this file will be cumbersome. Or they can always just use a different ip address temporarily to allow ssh to the nodes.

[tserong]

I'm not sure we can write the 91_networkmanager file before upgrade -- it's generated by the harvester-installer binary from v1.7, which is available inside the upgrade container, but not really anywhere else. That file gets generated when upgrade_node.sh runs, so it happens during upgrade, but before the node is rebooted and thus before the problem potentially occurs. So it will be there, and just needs that little tweak, but you're right, getting to the node when it's IP address has changed could be irritating...