The ScyllaDB team announces the release of Scylla Manager 2.2.1, a production-ready Scylla Manager patch release of the stable Scylla Manager 2.2 branch. As always, Scylla Manager customers and users are encouraged to upgrade to Scylla 2.2.1 in coordination with the Scylla support team.
Scylla Manager is a centralized cluster administration and recurrent tasks automation tool.
Note that this upgrade only affects the Manager Server and sctool, not the Manager Agent.
Related Links
- Download Scylla Manager and Scylla Manager Agent for Scylla Enterprise users
- Download Scylla Manager and Scylla Manager Agent for open source users (up to 5 nodes)
- Scylla Manager Docker instance and example
- Scylla Manager 2.2 release notes21
- Scylla Manager Docs documentation
- Upgrade from Scylla Manager 2.x.y to Scylla Manager 2.2.1
- Submit a ticket for questions or issues with Manager 2.1 (Scylla Enterprise users)
Scylla Manager 2.2.1 features the following changes.
Row-level repair improvements
Scylla Manager 2.2.1 improves performance and stability of row-level repair introduced in Scylla Manager 2.2.0. Now repair is split into more Scylla repair jobs, and we issue one job per shard. If one of the token ranges takes longer to repair it does not hold completion of other jobs. The below diagram shows parallel performance in Scylla Manager 2.2.1 vs 2.2.0 in our test running repairs under load. The benchmark is described in our earlier blog post Repair Revisited.
Repair long polling
Scylla Manager now supports a repair long polling API added in Scylla 4.3. For that version we issue one request every 10 seconds to check Scylla repair job status. Also, we get the result immediately when the job ends.
Docs have a new home
Our Scylla Manager documentation has been refined and is now served from a new site: https://scylladb.github.io/scylla-manager.
21 Dec 2020