v1.6.0

Release date: 2026-06-02

Highlights

Added Large File Support (LFS) across the proxy, console, operator surface, and multi-language client SDKs. This release introduces a new LFS HTTP API, LFS tracker flow, console visibility, and deployment assets for running the feature in Kubernetes and Docker-based environments.
Added DR-side topic recovery with kafscale-cli restore, including exact restore handling when segment contents permit safe truncation at a timestamp boundary.
Improved Kafka compatibility and runtime behavior with on-demand metadata refresh, broader LIST_GROUPS version negotiation for older admin clients, group coordination fixes, and processor durability work for restore and parallel processing paths.

Helm users must treat this as an upgrade with manual review. The standalone lfs-proxy values surface was removed. LFS now hangs off the unified proxy and is enabled through proxy.lfs.enabled=true.
Helm defaults changed in ways that can affect existing values overlays. operator.replicaCount now defaults to 1, and several auth and etcd password fields now expect null or existingSecret wiring instead of inline default strings.
Broker examples and restore flows were hardened to avoid unsafe S3 defaults. Review any local demo or example-based S3 configuration before upgrading.

Fixed Kafka admin compatibility for consumer group listing by widening LIST_GROUPS version advertisement to support older clients.
Fixed broker behavior to return protocol errors instead of dropping connections in failing request paths.
Fixed duplicate offset assignment risk with lease-based partition ownership and reduced group coordination split-brain behavior by routing coordination requests through a single broker path.
Fixed restore and read paths to handle orphaned S3 segments gracefully and added exact-cutoff recovery behavior when segment structure makes it safe.
Fixed produce request encoding for multi-partition tagged fields and improved metadata routing and refresh behavior during fetch and produce operations.

Blocked S3 bucket takeover scenarios caused by unsafe example defaults in broker and LFS-related paths.
Bumped pgproto3 to a patched pseudo-version and merged follow-up dependency remediation across storage, controller, and workflow dependencies.

Replaced the hand-rolled Kafka wire protocol path with franz-go kmsg handling to reduce protocol drift and improve compatibility coverage.
Added an explicit CodeQL workflow for PR and push coverage and refreshed multiple CI, Docker, Go, and AWS dependency versions.
Expanded onboarding and example coverage with the E10-E50 walkthroughs and LFS demo assets.