隨著容器虛擬化技術的發展與成熟，微服務架構漸漸地吸引人們的注意，越來越多服務供應商採用容器工具來建構微服務系統，目前市面上最常以Kubernetes容器管理系統作為實現微服務架構的工具之一。微服務串鏈為微服務架構中最為常見的服務型態，而服務供應商為了滿足服務級別協議 (Service Level Agreement)，須確保服務有高可用性，所以在微服務架構中如何維護微服務串鏈的穩定，即可保證服務的可用性。雖然Kubernetes有原生的服務修復機制，然而Kubernetes對於節點錯誤的偵錯修復機制過於緩慢，因此本篇論文將針對節點錯誤，提出微服務串鏈高可用性機制讓微服務在節點發生軟硬體損壞時，高可用性機制可以進行錯誤偵測與服務修復動作來維持微服務串鏈的運作，並且我們利用實驗展示高可用性機制對比於Kubernetes原生機制能降低至少89%中斷時間。;With the development of container virtualization technology, microservice architecture gradually attracts people’s attention. More and more service providers build microservice systems with container tools. So far, Kubernetes is one of the most popular container tools for implementing microservice systems. To meet Service Level Agreement, service providers should ensure high availability of service. The microservice chain is the most common service type in microservice architecture, therefore, how to maintain microservice chain is the key point in the problem of high availability of service. Kubernetes has a native microservice recovery mechanism for error container, but the native microservice recovery mechanism for node failure takes lots of time. Thus, this paper proposed microservice-chain-based HA method for node failure. The proposed method can rapidly detect node failure, such as network failure, OS hang, or Poweroff, and recover the microservice. In addition, we also do the experiment and demonstrate that the proposed HA method can reduce the overall downtime at least by 89% compared with Kubernetes native recovery mechanism.
