Logical Router Component Migration via BFD State Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In managed virtualized networks, temporarily taking down a single edge node can affect multiple logical routers, necessitating techniques for individual management of centralized routing components without impacting others on the same host machine.
Innovation Solution
A method allowing administrators to force failover of a centralized routing component from one host machine to another with minimal packet and connectivity loss, using bidirectional forwarding detection (BFD) messages to manage the active-standby states of routing components and update their administrative states without affecting other components on the same host machine.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a physical host machine is temporarily taken down to migrate a centralized routing component, then the routing component can be migrated to a standby host, but other logical routers operating on the same host machine are negatively affected
Solution Approach 1:
The system segments the host machine's functionality by distinguishing between different logical router components. Instead of taking down the entire host machine, only the specific centralized routing component being migrated is deactivated. This allows other logical routers on the same host to continue operating independently, thus resolving the contradiction between enabling migration and avoiding impact on other routers.
Solution Approach 2:
The patent extracts the specific centralized routing component from the host machine context for migration purposes. By isolating the component to be migrated and separating its lifecycle management from the overall host machine status, the system can perform migration operations on individual components without affecting the host or other components, thereby eliminating the harmful impact on other logical routers.
2Object-affected harmful factors
If centralized routing components are individually managed without taking down the host machine, then other logical routers remain unaffected, but the complexity of managing active-standby states increases
Solution Approach 1:
The patent merges the state management of multiple centralized routing components into a unified system managed by the network controller. By combining the management of active and standby states across different components and hosts, the system provides a coordinated approach that handles failover and migration automatically, reducing the perceived complexity for individual component management while maintaining isolation between logical routers.
Solution Approach 2:
The system implements feedback mechanisms where the network controller monitors the status of centralized routing components and automatically adjusts active-standby states based on component availability and health. This automated feedback loop manages the complexity of state transitions, ensuring proper failover behavior without requiring manual intervention, thus balancing individual component management with system-wide coordination.
Data Source
AI summary
Some embodiments provide a method for a controller that manages a physical network that implements multiple logical networks that include multiple logical routers. The method receives a command to change a particular centralized routing component of a logical router to an inactive state. At least two centralized routing components of the logical router are implemented on at least two different host machines in the physical network. The method identifies a host machine on which the particular centralized routing component operates. Other centralized routing components of other logical routers also operate on the identified host machine. The method sends a message to the identified host machine to cause the particular centralized routing component to change to an inactive state, without modifying a state of the identified host machine or the other centralized routing components operating on the identified host machine.