[RELEASE] ScyllaDB Manager 3.6.0

Release Notes
, , ,
1

The ScyllaDB team announces the release of ScyllaDB Manager 3.6.0, a production-ready minor release of the stable 3.6 branch.

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

This release focuses on the new 1-1-restore procedure and support for the experimental native backup.

Below are the changes in this release.

1-1-restore

Scylla Manager 3.6.0 introduces a new sctool restore 1-1-restore command that significantly speeds up the restore process at the cost of stricter requirements:

  • 1-1-restore works only for vNode replicated keyspaces
  • Source and destination cluster must have mirrored topology:
    • The same number of nodes in each datacenter and rack
    • The same token assignment for each node

It sets restored tables and views `tombstone_gc’ mode to ‘repair’ which is required to avoid running repair operation as part of the restore procedure. After restoration, the ‘tombstone_gc’ mode can only be changed once the tables have been repaired - otherwise, data resurrection may occur.

Here is a benchmark of the regular (Load&Stream) restore and the 1-1-restore against a 3-node i4i.4xlarge cluster of ScyllaDB 2025.2.1 with a single table with 3TB of data after replication RF=3.

Chart1304×808 22.5 KB

Note that 1-1-restore performance is significantly improved starting from ScyllaDB 2025.2.1 thanks to the optimizations included in this patch release.

Native backup support

The new ScyllaDB 2025.2.0 release includes experimental support for native backup, which allows uploading sstables directly from ScyllaDB to S3 without ScyllaDB Manager Agent (rclone) proxy. Native backup proves to be much faster, but because of that, it might also impact online request latency. Its performance and impact can be controlled by setting stream_io_throughput_mb_per_sec in ‘scylla.yaml’.

The ScyllaDB Manager 3.6.0 release allows for experimenting with the native backup by setting the new sctool backup --method flag. It specifies the API used for uploading sstables:

  • auto: Use the native backup when possible, otherwise use the rclone.
  • native: ScyllaDB server uploads directly to backup location (supports only S3 provider).
  • rclone: ScyllaDB Manager Agent uploads to backup location.
    Since native backup is experimental, it is set to rclone by default.

All methods still require configuring the ScyllaDB Manager Agent with ‘scylla-manager-agent.yaml’. The native backup also requires object_storage_endpoints to be configured in ‘scylla.yaml’.

–rate-limit and –transfers flags do not take effect when using –method=native, as the upload performance is controlled by the ScyllaDB server (see stream_io_throughput_mb_per_sec in ‘scylla.yaml’).

Stability improvements

The ScyllaDB Manager 3.6.0 release also contains stability improvements:

  • Improve gocql port availability which caused errors on creating CQL session (#4469)

Upgrade to the new release

ScyllaDB customers are encouraged to upgrade to ScyllaDB Manager 3.6.0 in coordination with the ScyllaDB support team.

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

Useful Links:

Scylla Manager 3.6.0 supports the following ScyllaDB releases:

  • 2025.2
  • 2025.1
  • 2024.1

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

1 Like