[RELEASE] ScyllaDB Manager 3.13.0

The ScyllaDB team announces the release of ScyllaDB Manager 3.13.0, a production-ready minor release of the stable 3.13 branch.

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

This release focuses on stability and observability: it adds per-task metrics, reduces memory usage of ScyllaDB Manager Server and Agent, and fixes backups resumed after a pause or failure.

Below are the changes in this release.

Per-task metrics (#4917, #4925, #4926)

ScyllaDB Manager 3.13.0 adds metrics describing the state and progress of individual tasks, not only of the scheduler as a whole:

  • scylla_manager_scheduler_task_state

  • scylla_manager_scheduler_task_run_start_seconds

  • scylla_manager_scheduler_task_properties

  • scylla_manager_repair_task_properties

  • scylla_manager_backup_task_properties

  • scylla_manager_backup_task_progress

These metrics are the foundation for an upcoming ScyllaDB Manager dashboard in ScyllaDB Monitoring.

It also fixes the tablet repair progress metric, which never reached 100% for repaired tables, and restores the -2 (agent unreachable) status of the REST health check metric.

Reduced memory usage of ScyllaDB Manager Server and Agent (#4916, #4919, #4937)

ScyllaDB Manager 3.13.0 improves memory handling of the backup task, most notably:

  • Failed or retried backups - memory used for listing snapshot files is now released when listing times out, so ScyllaDB Manager Server memory usage no longer grows with each failed backup run. On clusters with many SSTables this growth could exhaust the memory of the ScyllaDB Manager node.

  • Backup deduplication - remote SSTables are no longer all loaded into memory before deduplication, which significantly reduces peak memory usage of the backup task.

  • ScyllaDB Manager Agent - listing snapshot directories no longer creates a new cached file system for each backup run, which keeps Agent memory usage bounded.

Resumed backup may miss the schema file (#4947)

A backup task run resumed after being paused or failing at an early stage could finish successfully without the schema file, so it could not be used for restoring the schema.

ScyllaDB Manager 3.13.0 addresses this issue by starting such runs over with a new snapshot. Snapshots of runs which won’t be resumed are now deleted right away (#4586).

Removed TLS 1.0 support (#4823)

TLS 1.0 is no longer supported. The tls_version option in ScyllaDB Manager Server and Agent configuration accepts only TLSv1.2 (default) and TLSv1.3.

Configurations still setting TLSv1.0 need to be updated before the upgrade.

Bug fixes and other improvements

  • close no longer needed ScyllaDB API clients and CQL sessions, e.g., of deleted clusters or after restore errors (#4922, #4936)

  • release scheduler resources and metrics of deleted tasks, and stop a running task when it is deleted (#4923)

  • fix filtering of ScyllaDB metrics, which caused querying all node metrics and frequent timeouts (#4921)

  • bound the number of ScyllaDB Manager Agent HTTP metric series (#4935)

  • lower initial schema agreement wait during restore from 15s to 1s (#4930)

  • back up, repair and restore new ScyllaDB 2026.4 alternator export tables (#4948)

  • dependencies upgrades (#4910, #4915, #4933)

Upgrade to the new release

ScyllaDB customers are encouraged to upgrade to ScyllaDB Manager 3.13.0 in coordination with the ScyllaDB support team.

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

Useful Links:

ScyllaDB Manager 3.13.0 supports the following ScyllaDB releases

  • 2026.3

  • 2026.2

  • 2026.1

  • 2025.1

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