A new console, IRSA for EKS, and Spinifex on Vultr
The console has been redesigned from the ground up, with every resource opening in a side panel and a new view of each node's hardware and health. EKS pods can now take an IAM role through their ServiceAccount, and Spinifex installs on Oracle Linux 10.2 hosts and on Vultr. Across EKS, ECS, ECR, load balancing, networking and RDS, responses and errors came further into line with AWS.
Console
- A complete redesign, with a new design system and every resource opening in a side panel.
- Every resource list has bulk delete, a shared filter bar and inline tag editing, and the EC2 instance, RDS database and Route 53 record forms have been reworked.
- A new nodes view shows hardware inventory, live host metrics, cluster topology and node-to-node network tests.
Platform
- Oracle Linux 10.2 is supported as a host operating system, from a single node to a multi-node cluster.
- Spinifex runs on Vultr, with a deploy script that stands up single or multi-node clusters on Vultr bare-metal or VM hosts.
- Package upgrades no longer restart the Spinifex, OVS and OVN services, and NetworkManager no longer replaces the host's systemd-resolved stub.
Kubernetes (EKS)
- IRSA: pods using an annotated ServiceAccount get that IAM role's credentials automatically, and trust policies accept AWS-format OIDC condition keys.
-
Each cluster gets a single AWS-style cluster security group, and
DeleteClusterreturnsDELETINGright away and finishes the teardown in the background. - Workers launch with an IMDS hop limit of 2, pull images from any node's ECR registry, and leave the cluster's Node list when terminated.
Load balancing and containers (ELBv2, ECS and ECR)
-
ALBs set
X-Forwarded-For,X-Forwarded-ProtoandX-Forwarded-Port, stop sending traffic to unhealthy targets, and health-check with HTTP/1.1, aHostheader and theELB-HealthCheckerUser-Agent. - ECR serves the OCI referrers API, so signatures and SBOMs attached to an image can be discovered, and its auth tokens stay valid after the role session that issued them expires.
-
ECS task definitions return AWS's defaults and registration timestamps, and
DescribeTasksreturnscreatedAt,startedAtandstoppedAt.
Security
- Changes to an account's status take effect immediately instead of waiting out a cache.
-
ECS task role credentials are served under a secret credential ID instead of the public task ID, and
iam:PassRoleis checked the same way whenever ECS or EKS takes a role. - ALB fixed-response bodies are rendered literally, and EBS encryption by default reports the platform's real key posture and refuses to be disabled.
Networking and reliability
- TCP connections no longer stall through the VPC gateway, and tenants with overlapping VPC CIDRs keep their public IPs across reconciles.
- Volume deletion is asynchronous, so a large volume can no longer fail its own delete.
- Account deletions resume automatically if the node running them fails, and a rebooted instance relaunches cleanly after the guest shuts down.