Container Cluster Update Control With Rollback and Minimal Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing container-based cluster update methods, such as rolling updates, do not allow for reversal in case of issues during the update process, while blue-green methods require double the resources.
Innovation Solution
A method that constructs a new container-based cluster with a minimum number of required nodes, migrates applications sequentially, and checks each step for normalcy, allowing rollback if necessary, using a load balancer and update control device to manage the process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If rolling update method is used, then update can be executed sequentially with operating system, but inability to return to previous state occurs when problem arises during update
Solution Approach 1:
The patent creates a backup of the current working cluster before starting the update process. This preliminary action stores the original state data (cluster configuration, application states, etc.) so that if problems occur during update, the system can revert to the backup state, thus resolving the contradiction between update efficiency and rollback capability
Solution Approach 2:
The patent creates a copy of the current cluster state as a backup before execution of update operations. This copying mechanism preserves the original state without interfering with the update process, enabling rollback capability while maintaining update productivity
2Reliability
If blue-green method is used, then rollback capability is provided, but resource consumption doubles during update work
Solution Approach 1:
The patent extracts only the essential state data (configuration files, application states, metadata) from the original cluster to create a compact backup, rather than maintaining a complete duplicate cluster environment. This extraction approach provides rollback capability while significantly reducing resource consumption compared to full blue-green deployment
Solution Approach 2:
The patent uses a lightweight backup mechanism that stores essential state information in a compact, disposable format. This backup can be quickly created and discarded or restored as needed, providing rollback capability without the heavy resource overhead of maintaining parallel cluster environments
3Productivity
If application is migrated to new cluster, then update can be executed, but irreversible process occurs in rolling update method
Solution Approach 1:
The patent performs preliminary backup of the original cluster state before initiating application migration to the new cluster. This preliminary action ensures that even as applications are migrated irreversibly to the new cluster, the original state is preserved in the backup, maintaining state reversibility while enabling update execution
Data Source
AI summary
An update control device of a container-based cluster update system constructs a master node in a new container-based cluster and constructs a minimum number of worker nodes required to execute a migration target application. The update control device deletes a migration target application of a container-based cluster in an initial state, deletes a worker node in which no application is installed, adds the worker node to a new container-based cluster, and sequentially migrates the migration target application to perform update.