Skip to content

Releases: cloudivian-org/AethelDB

Release list

AethelDB v0.2.0

AethelDB v0.2.0 Latest
Latest

Choose a tag to compare

View all tags
@vikasj2in vikasj2in released this 21 Jun 07:37
v0.2.0
afbd6f8
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

v0.2.0 — the operational-layer release. Builds on the 0.1.0 data plane (real WAL decode/redo, instant branching/PITR, quorum-replicated safekeepers, S3 offload, scale-to-zero proxy with TLS+SCRAM) with everything needed to actually operate it.

New in 0.2.0

  • Multi-tenancy — one page server hosts many fully-isolated tenants; reads and the control planes route by TenantId, tenants provisioned on first reference.
  • Durable tenant catalog — the tenant/timeline topology (with branch ancestry) is persisted to the object store and restored on restart.
  • Optional PgBouncer pooling — a verified, opt-in transaction-pooling tier (Docker and Kubernetes fixtures included).
  • Control-plane auth — optional --control-token gates the line + HTTP control planes (/healthz stays open).
  • Observability — a ready-to-run Prometheus + Grafana stack with a pre-provisioned AethelDB Overview dashboard, plus optional OpenTelemetry/OTLP trace export (feature-gated; default build unaffected).
  • WAL/page correctness — runtime --wal-redo backend selection; decode of pglz/lz4/zstd compressed full-page images; proxy CancelRequest routing.
  • Project hygiene — CI, expanded e2e suite, security hardening guide, and OSS community files.

147 tests pass with zero warnings. See CHANGELOG.md for the full list and docs/design/ for the subsystem designs.

Assets 2
Loading

AethelDB v0.1.0

AethelDB v0.1.0

Choose a tag to compare

View all tags
@vikasj2in vikasj2in released this 21 Jun 03:31
v0.1.0
c0e95ee
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

The first release of AethelDB — a working, end-to-end-tested serverless-PostgreSQL data plane. Compute (a stateless PostgreSQL engine) is separated from storage (a virtual, log-structured page layer): databases scale to zero when idle, cold-start on the next connection, branch their history instantly, and time-travel to any past point.

Highlights

  • Real WAL decode + redo through a verified postgres --wal-redo core mode.
  • Instant branching & point-in-time reads (copy-on-write across the ancestor chain).
  • Quorum-replicated safekeepers with real network replication + leader election.
  • Compaction + branch-aware GC, S3/MinIO offload.
  • Proxy: scale-to-zero activation, TLS termination, SCRAM-SHA-256 auth, and an opt-in Kubernetes-native activator (verified against a real cluster).
  • Ops: Prometheus metrics, HTTP/JSON control plane, Docker Compose + Kubernetes manifests, CI, and a Python e2e suite.

See CHANGELOG.md for the full list and docs/design/ for the subsystem designs.

Loading