Skip to content

CHANGELOG.md

Latest commit

 

History

History
219 lines (157 loc) · 11.4 KB

CHANGELOG.md

File metadata and controls

219 lines (157 loc) · 11.4 KB

Changelog

All notable changes to this project will be documented in this file.

[v0.3.1] - 2026-04-02

Bug Fixes

SSH Reliability

  • fix: add SSH keepalive and handshake timeout (#772) — SSH connections now send keepalive probes every 30 seconds to prevent session drops during long operations (e.g., kubeadm init). A 15-second handshake timeout prevents connectOrDie from blocking indefinitely against hosts that accept TCP but never complete the SSH handshake.

AWS Resource Cleanup — Provider

  • fix: HA NLB hairpin routing (#746, #762) — Control-plane nodes now use localhost:6443 for kubectl instead of the NLB endpoint, avoiding AWS NLB hairpin/loopback timeouts.
  • fix: switch HA NLB to internal scheme (#760) — NLB uses internal scheme to keep traffic within the VPC.
  • fix: handle InvalidInternetGatewayID.NotFound in IGW detach (#772) — The detach step now recognizes InvalidInternetGatewayID.NotFound alongside Gateway.NotAttached and skips retries.
  • fix: handle NotFound errors in NLB/listener/target-group deletion (#772) — All NLB cleanup paths now check for LoadBalancerNotFound, ListenerNotFound, and TargetGroupNotFound, treating already-deleted resources as success.

AWS Resource Cleanup — Periodic Cleanup Action

  • fix: NLB cleanup in periodic VPC cleaner (#762) — DeleteVPCResources now deletes NLB listeners, target groups, and load balancers before attempting subnet/IGW/VPC deletion, preventing DependencyViolation errors from NLB-owned ENIs.
  • fix: revoke cross-referencing SG rules before deletion (#766) — Security groups that reference each other are now cleaned up by revoking all ingress/egress rules before attempting deletion.
  • fix: treat InvalidVpcID.NotFound as success in VPC cleanup (#769) — VPCs that no longer exist are treated as successfully cleaned up.
  • fix: suppress NotFound warnings in all cleanup delete functions (#772) — The periodic cleanup job no longer logs misleading warnings when IGWs, security groups, subnets, or route tables are already gone.

CI

  • ci: update periodic cleanup and add manual trigger (#758, #765) — Periodic cleanup workflow uses the latest holodeck binary and supports manual dispatch.

[v0.3.0] - 2026-03-30

Features

Production-Grade Cluster Networking (#720–#728)

A complete overhaul of multi-node cluster networking for production workloads:

  • Add cluster networking cache constants and struct fields (#720, @ArangoGutierrez)
  • Add public subnet for cluster mode (#723, @ArangoGutierrez)
  • Add Transport abstraction for SSH connections (#724, @ArangoGutierrez)
  • Separate control-plane and worker security groups (#725, @ArangoGutierrez)
  • Cleanup dual security groups on cluster delete (#726, @ArangoGutierrez)
  • Wire SSM transport for private-subnet cluster nodes (#727, @ArangoGutierrez)
  • Wire cluster networking — private subnets, NAT, hostForNode, real tests (#728, @ArangoGutierrez)
  • Add ELBv2/NLB support for HA clusters (#614, @ArangoGutierrez)

Custom Templates (#701–#706)

User-defined provisioning templates with full lifecycle phase support:

  • Add CustomTemplate type and TemplatePhase enum to API (#701, @ArangoGutierrez)
  • Implement custom template loader and executor (#702, @ArangoGutierrez)
  • Add custom template input validation (#703, @ArangoGutierrez)
  • Integrate custom templates into dependency resolver (#706, @ArangoGutierrez)

RPM Support (#676–#681, #693)

First-class support for RPM-based Linux distributions including Rocky Linux 9, Amazon Linux 2023, and Fedora 42 across all runtime stacks (Docker, containerd, CRI-O):

  • Add RPM support to Docker template (#677, @ArangoGutierrez)
  • Add RPM support to NVIDIA driver template (#676, @ArangoGutierrez)
  • Add RPM support to Kubernetes templates (#678, @ArangoGutierrez)
  • Add RPM support to kernel template (#681, @ArangoGutierrez)
  • Add RPM support to CRI-O template (#680, @ArangoGutierrez)
  • Add RPM support to container-toolkit package template (#679, @ArangoGutierrez)
  • Add RPM docs, e2e validation, and Fedora 42 support (#693, @ArangoGutierrez)
  • Add DNF/YUM package manager support to provisioner

ARM64 Support

  • Add AMI architecture detection and cross-validation (#664, @ArangoGutierrez)
  • Infer AMI architecture from instance type for ARM64 support (#669, @ArangoGutierrez)
  • Propagate image architecture in cluster mode (#661, @ArangoGutierrez)
  • Add ARM64 GPU end-to-end test on merge to main (#670, @ArangoGutierrez)
  • Detect Kubernetes arch at runtime instead of defaulting to amd64 (#663, @ArangoGutierrez)
  • Use runtime arch for NVIDIA CUDA repository URL (#662, @ArangoGutierrez)
  • Make architecture field case-insensitive for backward compatibility

Multi-Node Kubernetes Cluster Support (#562, #660)

  • Add full multinode Kubernetes cluster support (#562, @ArangoGutierrez)
  • Parallelize node provisioning, join info, and source/dest check (#660, @ArangoGutierrez)
  • Fix cluster-mode OS resolution to use per-node specs

Multi-Source Installation (#635–#637)

  • Add component provenance tracking to environment status (#635, @ArangoGutierrez)
  • Support multiple installation sources for container runtimes (package, runfile, git) (#637, @ArangoGutierrez)
  • Support multiple installation sources for NVIDIA drivers (package, runfile, git) (#636, @ArangoGutierrez)
  • Add git ref resolution for CTK installation from GitHub sources
  • Add Kubernetes installation from custom sources

CLI Improvements (#563, #621)

  • Complete CLI with full CRUD operations (create, delete, list, status, dryrun) (#563, #621, @ArangoGutierrez)

AWS Infrastructure

  • Add retry logic with exponential backoff (#616, @ArangoGutierrez)
  • Add cleanup mode for standalone VPC cleanup to GitHub Action (#4938a2ee, @ArangoGutierrez)
  • Replace bash cleanup script with native Go implementation (@ArangoGutierrez)
  • Idempotent provisioning templates with enhanced error handling (#570, @ArangoGutierrez)
  • Add Ubuntu 20.04 to OS AMI registry

Stale Issue Management (#695)

  • Add GitHub workflow to mark issues as stale after 90 days of inactivity (#695, @ArangoGutierrez)

Bug Fixes

Cluster Networking

  • Wait for NAT Gateway available state before creating routes, fixing race condition (#735, @ArangoGutierrez)
  • Verify API server against local IP before switching to NLB (#721, @ArangoGutierrez)
  • Remove local kubeadm config cleanup that races in cluster mode (#718, @ArangoGutierrez)
  • Make kubeadm config local path unique per environment (#717, @ArangoGutierrez)
  • Guard substring slice and redact join credentials in logs (#654, @ArangoGutierrez)
  • Surface NLB errors instead of swallowing them (#645, @ArangoGutierrez)
  • Disable Source/Dest Check on ENI for single-node deployments
  • Restrict security group CIDR to detected public IP (#615, @ArangoGutierrez)
  • Fail instead of silently falling back to 0.0.0.0/0 for security group (#650, @ArangoGutierrez)
  • Implement cleanup on partial creation failures (#612, @ArangoGutierrez)
  • Propagate image architecture in cluster mode (#661)

Provisioner / SSH

  • Close SSH client before reassign, close pipe reader (#659, @ArangoGutierrez)
  • Split SSH sessions in createKindConfig (#657, @ArangoGutierrez)
  • Remove unreachable code in connectOrDie, wrap last error (#655, @ArangoGutierrez)
  • Correct retry count in connectOrDie error message (#628, @ArangoGutierrez)
  • Guard nil provider for SSH provider mode (#643, @ArangoGutierrez)
  • Close SSH client in GetKubeConfig (#640, @ArangoGutierrez)
  • Use TOFU known_hosts for interactive SSH sessions (#653, @ArangoGutierrez)
  • Add mutex to TOFU known_hosts to prevent race condition (#644, @ArangoGutierrez)
  • Use TOFU host key verification in all SSH connections (#625, #630, @ArangoGutierrez)
  • Log SFTP MkdirAll error instead of discarding (#642, @ArangoGutierrez)

Security

  • Validate node labels and IPs before shell interpolation (#656, @ArangoGutierrez)
  • Security and concurrency: template input validation, error wrapping (#623, @ArangoGutierrez)
  • Critical bugs: storageSizeGB race, log.Fatalf in goroutine, file permissions (#622, @ArangoGutierrez)
  • Create kubeconfig with 0600 permissions (#629, @ArangoGutierrez)
  • Handle crypto/rand failure in retry jitter (#641, @ArangoGutierrez)
  • Validate instance type before creating resources (#672, @ArangoGutierrez)

AWS Provider

  • Preserve error chain with errors.Join, copy tags for goroutines (#651, @ArangoGutierrez)
  • Return errors from GenerateInstanceID, validate ID format (#648, @ArangoGutierrez)
  • Replace context.TODO() with proper timeouts in create.go (#611, @ArangoGutierrez)
  • Add context propagation and timeout support to cleanup (#608, @ArangoGutierrez)
  • Address ignored errors with logging and documentation (#610, @ArangoGutierrez)
  • Use %w for proper error wrapping (#609, @ArangoGutierrez)
  • Validate instance type before creating resources (#672)

Templates and Provisioning

  • Pre-release validation fixes for kubeadm, networking, and CTK (#716, @ArangoGutierrez)
  • E2E fixes for RPM-based distros (Rocky Linux 9)
  • Validate feature gates, track branches, and endpoint host (#647, @ArangoGutierrez)
  • Address audit fixes for heterogeneous cluster support
  • Restore CTK/K8s validation and fix test schema

Logger / Utility

  • Replace shared channels with per-invocation context cancellation (#631, @ArangoGutierrez)
  • Nil out completed cancel entries in activeCancels (#646, @ArangoGutierrez)
  • Replace signal goroutine with signal.NotifyContext (#649, @ArangoGutierrez)

Tests

  • Use valid hex instance IDs in delete and status tests (#713, @ArangoGutierrez)

Performance

  • Inject sleep function to speed up AWS provider tests (~200x faster) (#739, @ArangoGutierrez)
  • Parallelize node provisioning, join info, and source/dest check in cluster mode (#660, @ArangoGutierrez)
  • Pre-compile templates, hoist regex, filter DescribeLoadBalancers (#652, @ArangoGutierrez)

CI/CD

  • Split E2E tests into smoke (pre-merge) and full (post-merge) tiers (#740, @ArangoGutierrez)
  • Add ARM64 GPU end-to-end test on merge to main (#670, @ArangoGutierrez)
  • Add workflow to mark issues as stale with no activity (#695, @ArangoGutierrez)
  • Pin holodeck action to v0.2.18 in periodic CI workflow

Documentation

  • Update multinode cluster guide with production-grade networking (#734, @ArangoGutierrez)
  • Comprehensive documentation refresh (#707, @ArangoGutierrez)
  • Add multi-source installation guides and examples (#634, @ArangoGutierrez)
  • Add OS selection guide
  • Add CTK multi-source installation guide
  • Add GitHub project management templates and AI instructions

Dependencies

  • Updated AWS SDK Go v2 to latest (multiple bumps: aws-sdk-go-v2, service/ec2, service/ssm, service/elasticloadbalancingv2, config)
  • Updated Kubernetes libraries: k8s.io/apimachinery, sigs.k8s.io/controller-runtime to v0.23.3
  • Updated golang.org/x/crypto to v0.49.0 and golang.org/x/sync to v0.20.0
  • Updated Go base image from 1.25.5 to 1.26.1 (bookworm)
  • Updated GitHub Actions: docker/build-push-action v7, docker/setup-buildx-action v4, docker/metadata-action v6, docker/login-action v4, docker/setup-qemu-action v4, aws-actions/configure-aws-credentials v6, actions/upload-artifact v7
  • Updated github.com/onsi/ginkgo/v2 and github.com/onsi/gomega to latest
  • Updated github.com/pkg/sftp to v1.13.10

[v0.2.18] - 2025-12-13

See GitHub Release