Version 3.8.4 of Eclipse Ditto, released on 07.11.2025
Edit this page

This is a bugfix release, no new features since 3.8.3 were added.

Changelog

Compared to the latest release 3.8.3, the following changes and bugfixes were added.

Bugfixes

This is a complete list of the merged pull requests.

Policies with expired (but not yet deleted) subjects could not be updated

Issue #2233 described a bug where policies could not be updated when policy subjects in that policy were expired and failed to be deleted from the policy. This was fixed in PR #2253.

Tune CBOR serialization and JsonSchema validation

PR #2252 contains performance improvements for Ditto-internal CBOR serialization and JsonSchema validation, reducing the amount of bytes allocated on the heap.

Optimize communication for thing/policy/connection existance checking by bypassing sharding proxy

PR #2254 bypassed the “sharding proxy” in order to check for existence of things, policies and connections. This reduces the communication overhead and improves reliability in case shards are being moved in the cluster, e.g. during a rolling restart.

Fix configuring missing cluster sharding healthcheck names

PR #2255 adds healtcheck configuration for Ditto shard region names to ensure that only pods which allocated shard regions are considered healthy by Kubernetes readiness probes.
This should improve reliability of Ditto clusters during rolling restarts.

Fixed exceptions during policy enforcer loading being silently treated as “empty enforcer”

PR #2256 fixes an issue where exceptions during loading of policy enforcers were silently treated as “empty enforcer”, leading to unexpected authorization results on e.g. “Ask timeout” errors, especially during rolling restarts.

Helm chart

Enhanced Helm configuration with terminationGracePeriodSeconds, garbage collection logging and more

PR #2258 enhances the available Helm configuration options with:

  • Kubernetes terminationGracePeriodSeconds configuration for Ditto pods
  • Garbage collection logging configuration
  • Pekko cluster coordinated shutdown phases configuration
  • “Ask with retry” configuration options for “edge” services (gateway, connectivity) and for enforcement (things, policies services)