Spinnaker Clouddriver pod 转到 CrashLoopBackOff

Question

通过 hal deploy apply 将 Spinnaker 部署到 EKS 时，Spinnaker Clouddriver pod 转到 CrashLoopBackOff 并出现以下错误，

Factory method 'awsProvider' threw exception; nested exception is java.lang.NullPointerException: Cannot get property 'name' on null object

我的 Halyard 配置如下，

currentDeployment: default
deploymentConfigurations:
- name: default
  version: 1.17.6
  providers:
    appengine:
      enabled: false
      accounts: []
    aws:
      enabled: true
      accounts:
      - name: my-account
        requiredGroupMembership: []
        providerVersion: V1
        permissions: {}
        accountId: '010101010101' # my account id here
        regions: []
        assumeRole: Spinnaker-Clouddriver-Role
        lifecycleHooks: []
      primaryAccount: my-account
      bakeryDefaults:
        baseImages: []
      defaultKeyPairTemplate: '{{name}}-keypair'
      defaultRegions:
      - name: us-east-1
      defaults:
        iamRole: BaseIAMRole

我的 Spinnaker-Clouddriver-Role IAM 角色目前拥有完全权限。我该如何解决这个问题？

这是完整的日志https://gist.github.com/agentmilindu/cfbebffe46b93458df8158f9355e4041

Answer 1

这或多或少是一种猜测，因为您没有提供关于三角帆设置的一丁点版本信息，但是...

根据at com.netflix.spinnaker.clouddriver.aws.provider.agent.ReservationReportCachingAgent$_determineVpcOnlyAccounts_closure2.doCall(ReservationReportCachingAgent.groovy:117) ~[clouddriver-aws.jar:na] in your gist, which corresponds to getAmazonEC2(credentials, credentials.regions[0].name) in version 6.5.2

看来他们不能像您一样容忍空的 regions: []；因此：

aws:
  enabled: true
  accounts:
  - name: my-account
    # ... snip ...
    # vvv-- update this list
    regions:
    - name: us-east-1

Spinnaker Clouddriver pod 转到 CrashLoopBackOff

Spinnaker Clouddriver pod goes to CrashLoopBackOff

amazon-web-services

kubernetes

spinnaker

spinnaker-halyard

amazon-eks