- Bare Metal support. Cluster node pools deploy on Virtual Instances or Bare Metal Kubernetes instances, including GPU Bare Metal instances, because dedicated hardware reduces latency and avoids noisy-neighbor effects from shared VMs.
- Secure master node management. Customers access worker nodes directly, while Gcore administrators manage master nodes and the control plane. Access to master nodes is restricted to maintain security and service stability.
- Flexible pool configuration. Each cluster organizes worker nodes into pools. Add pools and set Minimum nodes and Maximum nodes per pool on the Create cluster form in the Gcore Customer Portal.
- Autoscaling. The Cluster Autoscaler adds and removes nodes based on demand when pool minimum and maximum limits differ. Configure autoscaling in pool settings and in the Cluster Autoscaler section under Advanced settings.
- Autohealing nodes. The service monitors node health and replaces failed nodes automatically. The Autohealing nodes option is enabled by default when creating a pool.
- Data storage. PersistentVolumeClaims (PVC) work with Gcore volumes to persist application data. Create a PVC from the storage settings.
- Networking. Route application traffic with an Nginx ingress controller.
- High availability and SLA. Managed Kubernetes is built on Cluster API and includes a 99.99% SLA.
Shared responsibilities
Managed Kubernetes handles many operational tasks, but not all responsibilities for running applications on Kubernetes. Gcore follows a shared responsibility model. Gcore’s responsibilities:- Managing the control plane.
- Maintaining high availability and SLAs.
- Provisioning worker nodes according to cluster settings.
- Managing SSH keys according to industry best practices.
- Validating that services and applications work with selected settings and configurations.