Changes and Releases

Updates on Giant Swarm workload cluster releases, apps, UI improvements and documentation changes.

  • Breaking changes

    In older releases the NGINX IC LoadBalancer Service name was hardcoded to nginx-ingress-controller. As of this release, to ensure the Service name uniqueness for multiple NGINX ICs per cluster support, the LoadBalancer Service name was made to be dynamic, derived from Helm release i.e. App Custom Resource (CR) name. Therefore, if you’re upgrading from an older NGINX IC App release to v1.8.0+, existing NGINX IC LoadBalancer Service may get replaced by a new one for every NGINC IC App CR whose name is not nginx-ingress-controller. When NGINX IC LoadBalancer Service gets recreated, cloud service provider (CSP) load balancer behind it gets recycled as well. It can take minute or so for ingress DNS records to be updated by external-dns and change propagated to clients. During that time there’s ingress traffic downtime, since clients still resolve old no longer present CSP load balancer. Please take the potential ingress downtime (a minute or so) into consideration when planning the NGINX IC App upgrade from older to v1.8.0+. To make sure the downtime is shortest possible, external-dns availability is important precondition. In recent platform releases (Azure v12.0.2, and AWS v12.1.4 and v11.5.4) we’ve improved external-dns monitoring and alerting. Therefore, before upgrading NGINX IC optional app to v1.8.0+, please make sure that your cluster has been upgraded to the latest platform release.

    Added

    • Support multiple NGINX IC App installations per tenant cluster.

    Removed

    • Dropped support for deprecated configuration properties:
      • configmap.annotations-prefix
      • configmap.default-ssl-certificate
      • configmap.hpa-enabled
      • configmap.hpa-max-replicas
      • configmap.hpa-min-replicas
      • configmap.hpa-target-cpu-utilization-percentage
      • configmap.hpa-target-memory-utilization-percentage
      • configmap.ingress-class
  • Fixed

    • Prevent breaking the client’s kubeconfig if token renewal fails.

    Added

    • Add --use-alike-instance-types for node pools.
    • Update ‘Details on Kubernetes on Giant Swarm’ (#550)
    • Fix h5 top margin inside well (#552)
  • Added

    • Add AzureOperatorVersion to pkg/label.
    • Add MachinePool to pkg/label
    • Add MachinePoolName to pkg/annotation
    • Add ReleaseNotesURL to pkg/annotation.
    • Add descriptions to App, AppCatalog and Chart CRDs.
    • Add deprecation notice to ChartConfig CRD.

    Changed

    • Graduate AppCatalog CRDs to v1.
    • Graduate App CRDs to v1.
    • Graduate Chart CRDs to v1.
  • Added

    • Set default request/limits in deployment (cpu: 100m, memory: 100Mi) per best practice

    Changed

    • Sync with upstream
      • Bug fixes
      • Add namespace overwrite

    Updated

    • Update architect-orb to 0.10.0
    • kube-state-metrics chart 2.8.8 -> 2.8.11
    • grafana chart 5.1.4 -> 5.3.0
    • grafana chart requirement 5.1.* -> 5.3.*
    • configmapReloadImage 0.0.1 -> 0.3.0
    • hyperkubeImage 1.12.1 -> 1.16.12
  • Added

    • Set default request/limits in deployment (cpu: 100m, memory: 100Mi) per best practice

    Changed

    • Sync with upstream
      • Bug fixes
      • Add namespace overwrite

    Updated

    • Update architect-orb to 0.10.0
    • kube-state-metrics chart 2.8.8 -> 2.8.11
    • grafana chart 5.1.4 -> 5.3.0
    • grafana chart requirement 5.1.* -> 5.3.*
    • configmapReloadImage 0.0.1 -> 0.3.0
    • hyperkubeImage 1.12.1 -> 1.16.12
    • Upgrade cert-manager from 0.9.0 to 0.15.2 (#31)
      • This is a breaking change. Please review the upgrade notes here.
    • Upgrade helmclient to 1.0.2
    • Upgrade architect-orb to 0.10.0

    Added

    • Webhook component to validate requests and prevent incorrect configurations.
  • Added

    • Add CR templating for external use.
  • Changed

    • Add CRD metadata