Last week in scylladb.git master (issue #223; 2024-03-24)

This short report brings to light some interesting commits to scylladb.git master from the last week. Commits in the 7df3acd39c…6bdb456fad range are covered.

There were 149 non-merge commits from 24 authors in that period. Some notable commits:

The native nodetool status command now handles joining nodes better.

A reactor stall when reading the materializing very large schemas has been fixed.

During compaction tombstone garbage collection, we now skip the bloom filter checks if the sstable only contains young data.

During compaction tombstone garbage collection, we now consider the memtable only if it contains the key. This prevents old data in memtable from preventing tombstone garbage collection.

Tracking of memory within materialized view updates was improved.

The replace operation, when running with the same number of nodes as the replication factor (e.g. typically 3), the node replace operation now works correctly.

There is now documentation for initiating upgrade into consistent topology management and recovering from a disaster when it is enabled.

A crash when a view was created while a TRUNCATE operation is in progress was fixed.

The system.token_ring virtual table is now tablet-aware.

The native nodetool command now mimics Java-style command line handling for greater compatibility with scripts.

The native nodetool command can now query effective ownership on a table basis, useful with tablets.

Raft now handles lost quorums more flexibly, allowing callers to decide whether they want to timeout or not.

The service level subsystem moved its storage from the system_distributed keyspace to a new table system.service_levels_v2, managed by raft group 0. Data is automatically migrated.

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