Tags

23.06.0-275-g1e69a241b

23.06.0

[23.06.0] - 2023-06-16

Features

- Expose the Public Lua API `pico.*`. It's supplemented with a
  built-in reference manual, see `pico.help()`.

- _Compare and Swap_ (CaS) is an algorithm that allows to combine Raft
  operations with a predicate checking. It makes read-write access to
  global spaces serializable. See `pico.help('cas')`.

- _Clusterwide schema_ now allows to create global and sharded spaces.
  The content of global spaces is replicated to every instance in the
  cluster. In sharded spaces the chuncs of data (buckets) are
  distributed across different replicasets. See
  `pico.help('create_space')`.

- _Raft log compaction_ allows stripping old raft log entries in order
  to prevent its infinite growth. See `pico.help('raft_compact_log')`.

- Remove everything related to "migrations". They're superseeded with
  "clusterwide schema" mentioned above.

CLI

- `picodata run --script` command-line argument sets a path to a Lua
  script executed at startup.

- `picodata run --http-listen` command-line argument sets the [HTTP
  server](https://github.com/tarantool/http) listening address. If no
  value is provided, the server won't be initialized.

- `picodata connect` CLI command provides interactive Lua console.

Implementation details

- Picodata automatically demotes Raft voters that go offline and
  promotes a replacement. See docs/topology.md for more details.
  Replicaset leadership is switched too.

Compatibility

- The current version is NOT compatible with prior releases. It cannot
  be started with the old snapshots.

22.11.0

[22.11.0] - 2022-11-22

Features

- Brand new algorithm of cluster management based on _"governor"_ concept
  a centralized actor that operates topology and performs instances
  configuration.

- Instances state is denoted by the word _"grade"_ (current and target).
  The current grade of an instance is not what it tells about itself,
  but how other instances are configured with respect to the current
  one.

- Automatically configure _sharding_ (based on `vshard`) and perform
  rebalancing when replication factor is satisfied.

- Apply schema and data _migrations_ clusterwide.

- Shrink cluster by _expelling_ instances.

Compatibility

- Current version is NOT compatible with `22.07.0`. It cannot be started
  with the old snapshots.

22.07.0

[22.07.0] - 2022-07-08

Basic functionality

- Command line interface for cluster deployment.
- Dynamic topology configuration, cluster scaling by launching more instances.
- Support for setting the replication factor for the whole cluster.
- Failure domains-aware replicasets composition.
- Two kinds of storages:
  - based on Raft consensus algorithm (clusterwide),
  - based on Tarantool master-master async replication.
- Graceful instance shutdown.
- Automatic Raft group management (voters provision, Raft leader failover).
- Dead instance rebootstrap without data loss.
- Automatic peers discovery during initial cluster configuration.

0.1

0.1