Release notes

This page contains the release notes for KurrentDB v26.2.

26.2.3

19 October 2026

Subscriptions: Fixed error logs when a subscription ends because the node cannot serve it (PR #5801)

When a subscription ended because the node was not ready, too busy, or not the leader, the server logged an error with a full stack trace. It now logs a single warning with the reason, for example Subscription ... terminated: Server is not ready. Clients receive the same response as before.

26.2.2

7 October 2026

Fixed memory leak in gRPC streaming calls (PR #5764)

In 26.2.0 and 26.2.1, memory usage grew by a few bytes with each gRPC streaming call, such as streaming reads and subscriptions, and was not reclaimed when the call ended. Memory usage now remains stable.

Subscriptions: Added opt-in FellBehind notifications for gRPC clients (PR #5760)

When a live subscription cannot keep up and drops back to catch-up mode, the server can now send a FellBehind message to the client, with a timestamp and checkpoint, just as it already sends CaughtUp. Clients only receive it if they opt in by requesting compatibility level 2 on the subscription, so existing clients see no change.

26.2.1

5 October 2026

Schema Registry: Fixed queries missing recently registered schemas (PR #5758)

In 26.2.0, schema registry queries, such as listing schemas or checking compatibility, could miss schemas registered or changed while the node was running. When this happened, the logs showed an ArgumentNullException for sequenceId from DuckDBProjector. Queries now stay up to date, and after upgrading they also pick up any schemas they missed.

Scavenging: Fixed scavenge on older Linux distributions (PR #5756)

26.2.0 could not run scavenges on RHEL 8 or Ubuntu 20.04. The scavenge would immediately fail due to the version of glibc. This has been fixed.

26.2.0

30 September 2026

What's new

Find out what's new in this release.

Connectors: Reduced disk reads after leader election (PR #5743)

When a node becomes leader, it reads the log to determine which connectors should be running. Previously, it could reread a large portion of the log after each election, even when no connectors had changed or none were configured. The leader now records its progress regularly, so a new leader only reads recent events. The first election after upgrading may still read from the last recorded position.