Alauda Service Mesh

Configure and Validate Cross-Cluster Failover

This document explains how to achieve cross-cluster failover for applications based on a multi-cluster service mesh using the Bookinfo application as an example.

TOC

Steps

  1. In the left navigation bar, click Service Mesh > Mesh.

  2. Click the name of the service mesh to be configured.

  3. On the Mesh Policies tab, click Create Policy > Circuit Breaker on the right side of cluster c1.

    NOTE

    • If there are no special requirements, use the default parameters.

    • If a service is not configured with a Circuit Breaker Policy, the Failover configuration will not take effect for it, and ingress traffic will be randomly forwarded among service Pods across all clusters in the service mesh. If a service is configured with a Circuit Breaker Policy, ingress traffic will be routed within the service's cluster until all Pods in that cluster are tripped.

  4. On the Mesh Policies tab, click Create Policy > Regional Load Balancing on the right side of cluster c1.

  5. Select Failover as the Policy Type and configure the related parameters as described below.

    ParameterDescription
    Failure RegionThe region of the cluster where the failed service Pods are located, i.e., the region of the cluster where the policy is configured.
    Disaster Recovery RegionThe region that provides load balancing for the cluster where the failed service Pods are located.
    NOTE

    After configuring the failover policy, when the service Pods in the failure region cluster are tripped and the health ratio of service Pods in the cluster is less than 71.43% (100/Overprovisioning Factor), part of the ingress traffic for the service in that cluster will be prioritized to the service Pods in the disaster recovery region cluster.

    For example, after configuring a failover policy for cluster c1 (disaster recovery region is the region of cluster c2), if the health ratio of productpage service Pods in the failure region cluster c1 is 50%, then 50% * 1.4 = 70% of the traffic for the service in cluster c1 will be handled by the service Pods in the failure region cluster c1, and the remaining 30% of the traffic will be handled by the service Pods in the disaster recovery region cluster c2.
    If failover is not set, traffic will be randomly forwarded among clusters in the mesh.

    For details on traffic load priority, refer to Priority levels.

  6. Click Create.

Validation Method

  1. In the top navigation bar, click the product view switch to switch to Container Platform and enter the namespace ns1 under cluster c1.

  2. In the left navigation bar, click Compute Components > Deployments.

  3. Click the Deployment name of the productpage.

  4. Click Stop in the upper right corner to stop all Pods of the productpage.

  5. Refer to the Access Verification method of the productpage service to continuously access the productpage service based on the routing configuration (External Access Address) created by the gateway deployed in cluster c1.

    Note: If the Bookinfo application can still be accessed normally, it indicates that cross-cluster failover has been achieved. Additionally, you can go to the Service Mesh platform to further verify traffic forwarding by viewing the service topology (cross-cluster) and call chains.

ON THIS PAGE