CARD

CARD, centralized agent for redundancy in datacenter, is a mechanism designed from the bottom up to provide redundancy to virtual machines and docker instances using the hot-standby model and the virtual IP concept of VRRP.

Virtual machine or container instances are pared up into active-standby pairs. A centralized agent performs health check to both ends of the pair, and grants the virtual IP to the standby instance when the active instance is down. There are separated code for container and VM pairs. Service providers can record the instance pair information in the csv file, running the script once will trigger the health check of all pairs once. Periodic check can be implemented with cron job.

This small piece of code is a piece of the MIXXCLOUD project, as we wanted to implement high availability on our load balancer but faced issues running VRRP on container instances. This mechanism was a last minute walkaround to this problem.