Master upgrade Auto User + Manual User Node upgrade Auto + Surge upgrades User User Release channels yes no no Cluster Auto-scaling H no H Pod auto-scaling H + V no (metrics) H Node health- monitoring yes no (ASG) no (VMSS) Node auto-repair yes no (ASG) no (VMSS) Live migration yes no no
firewall rules, anti-spoofing Routable on GCP network (incl. Interconnect/VPN) Better performance & scalability Note: Service (ClusterIP) addresses are only available from within the cluster. - Intranode visibility (ΔAKS, ΔEKS) - No CoreDNS (ΔAKS, ΔEKS) - NodeLocal DNSCache (β, ΔAKS, ΔEKS)
only Instance Groups - not so great story - Autoneg GKE controller https://github.com/GoogleCloudPlatform/gke-autoneg-controller anthos.cft.dev/autoneg: '{"name":"ldn-2020-primary", "max_rate_per_endpoint":10}'
so great story so far kBatch (β, manual, AKS, ΔEKS) - High performance, high throughput and technical batch computing Config Sync (AKS, ΔEKS) - GitOps for cluster management Application Delivery (β, AKS, ΔEKS) - GitOps for application deployment