You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
New AWS-compatible managed Kubernetes service: full cluster lifecycle (CreateCluster -> ACTIVE, plus Describe/List/Delete) backed by a K3s control plane
Managed node groups (CreateNodegroup/UpdateNodegroupConfig/DeleteNodegroup) - workers auto-join the cluster on boot
IAM-authenticated kubectl: aws eks get-token resolves to AccessEntries and access policies, no aws-auth ConfigMap
IRSA: per-cluster OIDC provider with JWKS discovery, plus IAM OpenID Connect provider management
Managed add-ons API and public/private endpoint access (scoped by publicAccessCidrs)
Single spinifex-eks-node AMI for server and agent roles, installable with one spx admin images import
Identity (IAM/STS/IMDS)
IMDS v2: identity, metadata and IAM credentials over the standard link-local endpoint
STS: GetSessionToken for temporary session credentials
IAM: OpenID Connect provider CRUD for IRSA
Certificates (ACM)
New ACM-compatible service: ImportCertificate / DescribeCertificate / ListCertificates / DeleteCertificate for BYO certs, powering HTTPS load balancer listeners
Load Balancing (ELBv2)
Network Load Balancers: L4 data plane (TCP/UDP/TLS) on nginx with active per-target health checks
HTTPS listeners with TLS termination - ACM certificates and configurable SSL policies
New APIs: SetSubnets, SetSecurityGroups, SetIpAddressType, AddTags/RemoveTags, and listener-certificate management
Deletes are now idempotent on already-deleted resources (Terraform destroy-safe)
Resource Groups Tagging GetResources for AWS Load Balancer Controller discovery
Networking
ENI hot-plug now wires a real OVS tap with OVN binding - attach/detach a network interface on a running instance
Recycled Elastic IPs are reachable immediately (host neighbour priming + round-robin address allocation), so a just-released EIP no longer returns unreachable
Reliability
EKS cluster create/delete is crash-safe: partial-failure teardown leaks no NLBs, ENIs, or security groups, and reconcilers resume after a daemon restart
A control plane that never becomes healthy fails with a clear reason instead of hanging in CREATING
Load balancer microVM NIC fix (virtio_net dependency chain) and lb-agent nginx-start hang resolved - NLB targets no longer stall at 0/N