Traffic IP Groups and Fault Tolerance on GCE

This chapter describes how to implement front-end fault tolerance on Traffic Managers running in Google Compute Engine (GCE) through the use of External IP addresses and Traffic IP Groups. It should be read as an introduction to the capability and how it pertains to GCE in particular. For full details of fault tolerance and Traffic IP Groups, see the Virtual Traffic Manager: User’s Guide

The Traffic Manager can implement fault tolerance through the use of Traffic IP Groups and GCE External IP addresses. You configure a Traffic IP Group to contain two or more of the Traffic Managers in your cluster, together with one or more pre-reserved GCE External IP addresses. In a Traffic IP Group, such addresses are referred to as Traffic IP Addresses.

A Traffic IP Address is a permanent, externally-visible IP address that must remain highly available, and the Traffic Managers in the group share responsibility for keeping the Traffic IP Address raised regardless of individual instance failure. In other words, if a Traffic Manager that is currently receiving traffic over a Traffic IP Address suffers a failure, the remaining Traffic Managers in the group take on responsibility for maintaining the service and raise the relevant Traffic IP Address between themselves.

You can configure Traffic IP Groups on GCE as either active/active or active/passive. Furthermore, all Traffic IP Addresses hosted by the group are single-hosted. That is, responsibility for hosting Traffic IP Addresses is spread evenly between the Traffic Managers assigned to the group. Each Traffic IP Address is hosted on only one Traffic Manager, and transferred to another group member should the original host fail.