v0.3.0-beta3

What's Changed

• Add support for PERSIST journal mode by @benbjohnson in #168
• Add support for TRUNCATE journal mode by @benbjohnson in #169

Full Changelog: v0.3.0-beta2...v0.3.0-beta3

v0.3.0-beta2

What's Changed

• docs: disconnected fixup in store by @ryanrussell in #111
• Prefix FUSE debug logs by @benbjohnson in #113
• Use uppercase hex for node ID by @benbjohnson in #114
• WAL support by @benbjohnson in #120
• Add WAL checkpoint test by @benbjohnson in #121
• Fix return lock type for SHM lock by @benbjohnson in #123
• List journal filenames in root handle by @dangra in #125
• Updates workflows to support external PRs by @dangra in #126
• Force WAL fsync() by @benbjohnson in #127
• Remove dead code for WAL locking by @benbjohnson in #129
• Consolidate locks & guards by @benbjohnson in #130
• Fix snapshot in WAL mode by @benbjohnson in #131
• Continue running on error by @benbjohnson in #137
• Reduce logging by @benbjohnson in #143
• Rollback journal on startup by @benbjohnson in #144
• Validate that mount & data directories are different by @benbjohnson in #145
• Ensure mount directory exists before mounting by @benbjohnson in #146
• refactor: replace Readdir(-1) with os.ReadDir by @Juneezee in #147
• Fix locking when applying LTX file to replica in WAL mode by @benbjohnson in #148
• Retain OS page cache on open(2) by @benbjohnson in #150
• Attempt unmount for unclean exit by @benbjohnson in #151
• Specify 'exec' command from CLI by @benbjohnson in #152
• Add SHM locks to /debug/vars by @benbjohnson in #154
• Add endpoint for dynamically setting debug logging by @benbjohnson in #156
• Replay last LTX on recovery by @benbjohnson in #157
• Fix recovery on replication position mismatch by @benbjohnson in #158
• Clear LTX files on snapshot received by @benbjohnson in #160
• Remove pre-exec sleep by @benbjohnson in #161
• Print version & commit info on startup by @benbjohnson in #162
• Initialize DB.mode on open by @benbjohnson in #163

New Contributors

• @ryanrussell made their first contribution in #111
• @dangra made their first contribution in #125
• @Juneezee made their first contribution in #147

Full Changelog: v0.2.0...v0.3.0-beta2

v0.2.0

Overview

The v0.2.0 release primarily focuses on stability and hardening, however, it also includes a few important features. This release is also marks the end of the alpha phase of LiteFS and it is now considered "beta" software. It should generally run well, however, please report bugs to the GitHub repository.

Major features

Candidacy

By default, every node in the cluster can become the primary node. However, moving a primary from one region to another can change performance characteristics significantly so it may be useful to only allow certain nodes to become primary. You can now set the candidate flag in the config to false for any nodes which should not become primary:

candidate: false

Static Lease

LiteFS uses Consul to dynamically change the current primary in the cluster on the fly. However, for users that do not wish to run Consul, they can set a single fixed primary node. This is similar to how Litestream's read replication was meant to work. Please see #47 for more details.

Pull Requests

• Upgrade to LTX v0.2.0 in #45
• Implement candidacy in #46
• Implement static lease in #47
• Return ENOSYS for xattr requests by @darthShadow in #51
• Implement database snapshot to LTX in #53
• Update contribution policy in #54
• Retention enforcement in #57
• Implement statfs() callback in #58
• Support non-default page size in #59
• Expose DB position via FUSE in #60
• Optional registration of Consul node in #63
• Use default consul key, if not specified in #65
• Return hostname in primary file in #66
• Metrics in #67
• Update lock delay to 5s in #68
• Set hostname env variable, if unset in #69
• Default Advertise URL in #70
• Add configurable data directory in #71
• Rename pprof import in #72
• Fix database locks to use LockOwner in #74
• Atomically write LTX files on commit in #78
• Add unique, persistent node ID in #79
• Disconnect streams when primary status changes in #80
• Add DB nil check on HTTP server in #81
• Remove GITHUB_TOKEN & Makefile in #82
• Add staticcheck CI in #83
• Add errcheck to CI in #84
• Wait for ready state on startup in #85
• Ensure node does not startup if LTX files are invalid in #86
• Minor CI tweaks in #87
• Disable poll() in FUSE layer in #88
• Verify all LTX files on startup in #89
• Verify database checksum on startup in #90
• Fix locking on LTX apply in #91
• Verify txid/checksum on LTX frame in #92
• Add long-running, functional test in #93
• Fix http log formatting in #94
• Skip LTX checksum check on snapshot in #95
• Document 'exec' field in config in #97
• Fix LTX checksum on VACUUM in #98
• Refactor GitHub Actions in #99
• Mark ready after initial replication set in #100
• Fix FUSE return so replicas report a SQLITE_READONLY error in #101
• Remove DBID in #102
• Refactor RWMutex & RWMutexGuard in #103
• Move CI timeout to 'go test' command in #104
• Use HTTP/2 for internal endpoints in #106
• Require explicit data directory in #109
• Only release amd64 and arm64 in #110

New Contributors

• @darthShadow made their first contribution in #51

Full Changelog: v0.1.1...v0.2.0

v0.2.0-beta3

Only release amd64 and arm64

v0.1.1

This version of LiteFS replaces the FUSE implementation from github.com/hanwen/go-fuse to bazil.org/fuse. The new implementation seems to be much more stable on a variety of operating systems (Fedora, Alpine, GH Actions, etc).

Use of PRAGMA journal_mode = wal now returns an error since LiteFS does not yet support WAL mode. Also, additional test coverage has been added and increased testing will continue.

v0.1.0

This is the initial release of LiteFS. It is still in an alpha stage but the build is intended to make it easier to test.

v0.1.0-alpha3

Build on release

v0.1.0-alpha2

This release adds the .primary file handle to replica nodes. It can be read to find the current primary's advertise URL.

v0.1.0-alpha1

Fix streaming; wait to start subprocess