[RELEASE] ScyllaDB Manager 3.12.0

Release Notes
, ,
1

The ScyllaDB team announces the release of ScyllaDB Manager 3.12.0, a production-ready minor release of the stable 3.12 branch.

ScyllaDB Manager is a centralized cluster administration and recurrent tasks automation tool.

This release focuses on improvements to immutable backups on real-time archive GCS buckets.

Below are the changes in this release.

Immutable (WORM) backup based on event-based holds (#4885, #4891, #4894, #4899, #4900)

ScyllaDB Manager 3.11.0 introduced the immutable backup feature allowing for additional protection of backed up snapshot files.
ScyllaDB Manager 3.12.0 extends it with the new event-based-hold mode controlled with the –retention-lock-mode backup task flag.

This mode utilizes the following cloud provider features for protecting snapshot files:

  • Default bucket event-based holds - uploaded objects have an event-based hold set. This hold needs to be removed before a protected object can be deleted.
  • Bucket retention lock - removing event-based hold from an object starts specified retention period during which the object can’t be deleted.

ScyllaDB Manager 3.12.0 utilizes those bucket features in the following way:

  • Newly uploaded snapshot files automatically have event-based hold applied.
  • Deduplicated snapshot files keep their hold.
  • Files referenced by a previous snapshot which are not a part of the current one have their event-based holds released. This starts their retention period.

The main benefit of this approach is that it makes only a single per-file request and does not repeat those requests for deduplicated files.
When making highly deduplicated backups to a colder storage tier (low storage costs, high request costs), it’s possible that the request costs can dominate the overall backup costs.
The event-based-hold mode aims to significantly reduce costs in such scenarios.

The amount of requests setting/removing objects’ event-based holds as well as the total and deduplicated amount of snapshot files can be tracked with the following metrics:

  • scylla_manager_backup_set_event_based_holds
  • scylla_manager_backup_removed_event_based_holds
  • scylla_manager_backup_files_count
  • scylla_manager_backup_files_skipped_count

Note that as for the other –retention-lock-mode values, event-based-hold mode is supported for GCS backup provider only.
More details, limitations, and example usage can be found in the ScyllaDB Manager retention lock docs.

Skip CDC tables backup (#4875)

System CDC tables and CDC log tables contain short lived data that can’t be safely restored. Because of that, their restoration was always skipped during the restore task. Despite that, they were not automatically excluded by the backup task, which resulted in their snapshots occupying backup storage and incurring additional costs.

ScyllaDB Manager 3.12.0 addresses that by filtering out both system and log CDC tables during backup procedure, so that it is faster and cheaper.

Bug fixes and other improvements

  • fix resumed repair progress duration inflation (#4865)
  • harden jail escape checks for Rclone local backend (#4889)
  • don’t log errors when all CPUs are busy when pinning them during restore (#4893)
  • try parsing UUID of SSTables with unknown format during restore (#4884)

Upgrade to the new release

ScyllaDB customers are encouraged to upgrade to ScyllaDB Manager 3.12.0 in coordination with the ScyllaDB support team.

The new release includes upgrades of both ScyllaDB Manager Server and Agent.

Useful Links:

ScyllaDB Manager 3.12.0 supports the following ScyllaDB releases

  • 2026.2
  • 2026.1
  • 2025.1

You can install and run ScyllaDB Manager on Kubernetes using ScyllaDB Operator. More here.