Examples of additional recording bug kong

Examples of the container when the platform increases, K8S will play a new pod, the pod to the load balancing is not ip cache, resulting in a null pointer error java. The details have still, sometimes normal and sometimes error. The reason is kong cache update policy is only one work process to keep it updated, the lack of synchronization logic other work processes, woker after the update if the load is normal to fail otherwise. The solution: increased instances, coupled logic cache invalidation, and all work processes update the cache.