Releases: lni/dragonboat
Releases · lni/dragonboat
v3.1.3 released
Dragonboat v3.1.3 is a minor bug fix release. All v3.1.x users are recommended to upgrade.
Bug fixes
- fixed unbounded info log growth
v3.1.2 released
Dragonboat v3.1.2 is a minor bug fix release. All v3.1.x users are recommended to upgrade.
bug fixes
- fixed incorrect promotion of observer node.
v3.1.1 released
Dragonboat v3.1.1 is a minor bug fix release. All v3.1.0 users are recommended to upgrade.
Bug fixes
- Fixed a data race when requesting leader transfer
v3.1.0 released
Dragonboat v3.1 is a maintenance release with breaking change. All v3.0.x users are recommended to upgrade. Please make sure to carefully read the CHANGELOG below before upgrading.
Bug fixes
- Fixed ImportSnapshot.
New features
- Added NodeHostConfig.RaftEventListener to allow user applications to be notified for certain Raft events.
Improvements
- Made restarting an existing node faster.
- Immediately return ErrClusterNotReady when requests are dropped for not having a leader.
Breaking changes
- When upgrading to v3.1.x from v3.0.x, dragonboat requires all streamed or imported snapshots to have been applied. github.com/lni/dragonboat/tools/upgrade310 is provided to check that. See the godoc in github.com/lni/dragonboat/tools/upgrade310 for more details. For users who use NodeHost.RequestSnapshot to export snapshots for backup purposes, we recommend to re-generate all exported snapshots once upgraded to v3.1.x.
v3.0.2 released
- made membership change more resilient to unexpected requests (#75).
v3.0.1 released
- Fixed an error handling issue in LogDB when batching entries.
v3.0.0 released
New features
- Added on disk state machine (statemachine.IOnDiskStateMachine) support. The on disk state machine is close to the concept described in the section 5.2 of Diego Ongaro's Raft thesis.
- Added new API for requesting a snapshot to be created or exported.
- Added the ability to use exported snapshot to repair permanently damaged cluster that no longer has majority quorum.
- Added new API for cleaning up data and release disk spaces after a node is removed from its Raft cluster.
- Added the ability to limit peak memory usage when disk or network is slow.
- Added Go module support. Go 1.12 is required.
Improvements
- Further improved self checking on configurations.
- Added snapshot binary format version 2 with block base checksum.
- Synchronous variants have been provided for all asynchronous user request APIs in NodeHost.
Breaking changes
- The Drummer package has been made invisible from user applications.
- The statemachine.IStateMachine interface has been upgraded to reflect the fact that not all state machine data is stored in memory (#46).
v2.1.7
v2.1.6
- Removed Makefile targets for building drummer related server components.
- Updated drummer/README.md to mention the plan to move drummer into internal/drummer.
You can safely skip this minor release if you don't use the drummer package/have not heard about drummer.