Last week in scylladb.git master (issue #250; 2024-10-06)

This short report brings to light some interesting commits to scylladb.git master from the last week. Commits in the c17d353718…882a3c60e4 range are covered.

There were 77 non-merge commits from 23 authors in that period. Some notable commits:


After a tablet is migrated away from a node, a cleanup process takes place to remove the tablet’s footprint from the replica. During this process a race can happen the tablet’s compaction strategy is changed concurrently. This is now fixed.

The metric for the amount of memory used by commitog was updated incorrectly, indicating a leak that wasn’t there. This is now fixed.

A rare case where an sstable promoted index lookup would return in correct information was fixed.

The system.snapshots virtual table now lists snapshots of dropped tables.

The CREATE MATERIALIZED VIEW statement now supports the undocumented WITH ID clause, improving compatibility with Cassandra.

The nodetool replace and remove operations now deprecate using IP addresses to specify nodes. Use host IDs to specify nodes.

Repair has a read timeout in order to protect against deadlocks, but in practice the timeout caused repairs to fail. The timeout is now removed.

When a tablet is split into two, tombstone garbage collection becomes more complicated, since sstables for the tablet exist in both pre-split and post-split state. We are now more careful with tombstone garbage collection during this operation.

The restore REST API can now accept a list of sstables to restore.

Restore will keep sstables on object storage in place after restore.

The index page cache will now generate fewer disk IOPS if an index read is partially cached.


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