The ScyllaDB team announces ScyllaDB Open Source 5.0.9, a bugfix release of the ScyllaDB 5.0 stable branch. Like all past and future 5.x.y releases, it is backward compatible and supports rolling upgrades.
Note that the latest stable branch of ScyllaDB is 5.1, and you are encouraged to upgrade.
Related links:
- ScyllaDB Open Source 5.0
- Get ScyllaDB Open Source - AWS AMI, GCP Docker, binary packages and unified installer
- ScyllaDB Web Installer for Linux (all releases)
- Upgrade from ScyllaDB Open Source 5.x.y to Scylla Open Source 5.x.z
- Upgrade from ScyllaDB Open Source 4.6.x to 5.0.y
- Please let us know if you encounter any problems.
Issues fixed in this release:
- CQL: Different token calculation functions for “empty” partition keys cause sstable reader to misuse index #9352
- CQL: Wrong validation of blobs in INSERT JSON #10114
- CQL: batch statements, inserting a row with a null key column should be forbidden #12060
- CQL: Deleting a long base partition may leave some undeleted materialized view rows #12297
- CQL: in Scylla, returning incomplete results when using paging #12361 (introduced in 5.1.0 and 5.07). The issue manifests only in specific circumstances: when a page starts with a dead row. This page (and consequently the query) will be wrongfully terminated immediately. In 5.1 this has to be a row covered by the partition tombstone.
- CQL mutation::consume doesn’t emit a closing range_tombstone_change if the highest tombstone is open ended. In a rare use case, this can lead to missing data in the result. #10316
- Stability: Memtable(s) are not flushed when cleaning up a table, leaving disowned tokens in the memtable, which might be resurrected.#1239
- Stability: A crash in Scylla memory management code, triggered by index file caching, was fixed. The bug was caused by an allocation from within the memory allocator, causing cache eviction to free memory. Freeing the evicted items re-enters the memory allocator in a way that was not expected by the code. Fixes #9573
- Stability: preemption pitfall around waiting for readmission (found is fuzzy tests) #10187
- Stability: Gossip now uses a better source to get live node information for truncate, reducing false failures of the TRUNCATE operation. #10296, #11928
- Stability: During rebuild on asymmetric cluster several aborts and coredump happened #11923 (introduced by the fix for #11770 above)
- Init: scylla_setup “provisioning in a similar environment” message is missing a value for --online-discard parameter #11700
- Stability: reader_concurrency_semaphore: inactive readers are only evicted on the admission path #11770
- performance: parsers are compiled without inlining, even in release mode #12463