Last week in scylladb.git master (issue #194; 2023-08-27)

This short report brings to light some interesting commits to scylladb.git master from the last week. Commits in the 7275b8967c…9806bddf75 range are covered.

There were 86 non-merge commits from 17 authors in that period. Some notable commits:

Tablets are a new, experimental method of partitioning data among nodes, diverging from vnodes. Individual tablets are now stored in their own sstables, facilitating data movement and cleanup.

A bug that caused nodes to fail to start if a tablet was migrated concurrently with its table being dropped was fixed.

The system.tablets table stores the replica set of each tablet in the system. It now maintains the order of replicas, so that the notion of primary owner is maintained.

The scylla.yaml configuration items are now documented in the documentation website.

A recent regression causing a crash on table drop was fixed.

The setup utility supported an --online-discard switch to enable/disable online discard, but it did not actually work. This is now fixed.

The nodetool stop RESHAPE command is supposed to stop the reshape operation, but in fact only aborted running reshape compactions, which were promptly restarted. It now aborts the entire operation as expected.

A crash during rebuild operations with experimental consistent cluster topology was fixed.

ScyllaDB contains two classes of tables, system and user, and uses separate memory pools for their memtables. This avoids a deadlock when a user memtable is being flushed, and needs to allocate memtable space for a system table as part of the flush process. We now automatically designate all system tables as using the system memtable pool.

Latency during repair of large numbers of small rows was improved.

ScyllaDB now supports the –ignore-dead-nodes option family when experimental consistent cluster topology is enabled.

See you in the next issue of last week in scylladb.git master!