Lab testing
Ledger
Raft-replicated append service with checkpoint anchoring
Rust ledger service with HTTP append API, Raft consensus over UDP, persistent log state, and checkpoint/anchor workflows.
Current capabilities
- HTTP append endpoint for writing new records
- Raft-backed replication between cluster nodes
- Checkpoint records with cluster and hash metadata
- Config validation for node and peer setup
Not in scope yet
- A full query engine
- Dynamic peer discovery and automatic scaling
- General database-style read/write workflows
What ships on day one
- Leader-aware append endpoint with redirect responses for follower nodes
- Raft worker + UDP transport for replicated append commands
- Checkpoint records containing cluster id, committed index, and global hash
Deployment reality
Deploy as a cluster of nodes with one HTTP address and one Raft address per node.
- Self Hosted
Operational assumptions
- One node is elected leader for appends at a given time
- Followers redirect clients when they are not leader
- Peer addresses are fixed through configuration
- All nodes use the same cluster identity and genesis seed
Current boundaries
- Not intended as a relational or document database
- No broad query DSL in the current API
- No dynamic membership management in this release
Platform guarantees
Core behaviors in the current release.
Invariant 1
Writes are accepted through append operations; records are not edited in place.
Invariant 2
Leader-aware writes: followers return redirect or unavailable responses.
Invariant 3
Raft worker and UDP transport keep log replication active between peers.
Invariant 4
Checkpoint entries include cluster id, committed index, and global hash.
Invariant 5
Config validation prevents duplicate peer addresses and invalid checkpoint intervals.
Append API
Clients submit values to `/append` and receive committed index information.
Leader hints
Non-leader nodes return redirect responses to help clients find the leader.
Checkpoint metadata
Checkpoint payloads carry epoch, hash, and cluster details for downstream tooling.
Anchor directory
Checkpoint anchors can be written to a configured directory for external workflows.
API behavior
What clients should expect at runtime.
- Append returns record index and total committed count.
- Followers return redirects when leader location is known.
- If leader location is unknown, requests return service unavailable.
Operational readiness
Ready to shape your first cluster?
- Single binary service with clear HTTP + Raft entry points
- Config file supports explicit peer and checkpoint settings
- Dual-arch Linux binaries and OCI images are built from Bazel
Email labs@titaniumguard.in to plan cluster shape, checkpoint policy, and rollout steps.