This short report brings to light some interesting commits to scylladb.git master from the last week. Commits in the 755d528135c..f5112648316 range are covered.
There were 163 non-merge commits from 29 authors in that period. Some notable commits:
A new storage engine, logstor, has been merged. Logstor improves performance for key-value workloads and is intended for strongly consistent tables. It is now available in unsupported experimental mode only.
The Trie index is now the default sstable format for new clusters. Upgraded clusters retain the traditional index.
The Prometheus metric exporter now exports some systemd metrics.
A node decommission will now wait for a snapshot or backup to complete.
A replica will terminate a read request if it thinks it has low chance of completing before timeout. It is now careful to only terminate requests that haven’t been started yet; started requests have had some work invested into them so it’s worthwhile to try to let them complete.
Connections on the maintenance socket now have full authority on the database.
Read performance for workloads with very small partitions, using the traditional (non-Trie) index, has been greatly improved.
See you in the next issue of last week in scylladb.git master!