Scale operation on node pool causes existing nodes to restart

Products

• VMware Telco Cloud Automation

Issue/Introduction

You might see a behavior where a scale operation on the node pool(scale up or scale down) causes the existing nodes in that node pool to reboot. This a known issue  where any node pool operation triggers a NIC reordering. As a result, the nodeconfig operator interpret this as network change and will reboot the existing nodes. 

Environment

2.2

Resolution

TCA supports both early binding and late binding customizations for nodepools. Additional network interfaces can be configured to a nodepool either as early binding or as late binding. TCA has a merge logic to combine these early and late binding network interfaces. This merge logic doesn't retain the order of the network interfaces and hence the network interfaces reordering.
This merge logic is triggered whenever a nodepool edit operation is performed. And so in this case, at the time of scale-out, the network interface re-ordering was observed.

Workaround:
Create a nodepool without additional vmxnet3 network adapters(early binding), define all vmxnet3 network adapters in CSAR(late binding). Scale up (or any edit operation) of such a nodepool will not cause any re-ordering of network interfaces.

A fix for this issue is already available in TCA 2.3