Management cluster creation fails during the cluster control plane VM creation phase.
When observing the node and cluster logs, you will see the following sequence of errors:
API Server Connectivity Failures: The API server attempts to reach several Aggregated APIs but fails with the error: connect: no route to host
Network Interface Lease Drops: The systemd-networkd journal logs on the control plane node show that eth0 is constantly losing its DHCP lease, causing the bootstrap process to time out. You may see the following line repeating frequently (e.g., 24+ times): systemd-networkd[811]: eth0: DHCP lease lost
DHCP Server Errors: Checking the dhcpd journal logs on the DHCP server reveals IP lease duplication errors: client 00:00:00:00:00:00 has duplicate leases on 10.xx.xx.xx/24
TCA: 3.x, TCP: 5.x
Correlating the systemd-networkd and DHCP server logs identifies two concurrent issues causing the node to lose network connectivity during bootstrapping:
IP Duplication / Subnet Collision: There is a collision between the PXE boot environment and the TKGm CaaS environment sharing the same subnet/IP range.
Insufficient DHCP Lease Time: The default-lease-time on the DHCP server is set to 600s (10 minutes). This window is too short, forcing the bootstrap node to release its IP address before the cluster configuration process can successfully complete.
To resolve this issue and successfully deploy the control plane, apply both of the following infrastructure configuration changes:
Segregate IP Ranges: Use completely different subnets or dedicated IP ranges for the PXE boot environment and the TKGm CaaS environment to prevent lease duplication and IP collisions.
Increase DHCP Lease Time: Modify the DHCP server configuration to increase the default-lease-time to allow the bootstrap process enough time to complete.
Recommended Value: 86400s (24 hours)