HashiCorp Raft Chaos Testing

npm run stability:go-raft-chaos runs a HashiCorp Raft-focused stability harness. It is separate from normal E2E, Noosraft chaos, and soak tests.

The current harness reuses the real test:raft:go cluster path and records the covered HashiCorp Raft sidecar scenarios as a machine-readable report under go-raft-chaos-results/.

Covered now:

• baseline three-node HashiCorp Raft commit
• leader election and block replication
• leader sidecar restart/failover
• follower restart/catch-up
• persisted data directory restart
• forced sidecar death and TypeScript supervisor auto-restart
• TypeScript HTTP listener interruption while the HashiCorp Raft sidecar stays alive; missed committed callbacks are recovered by replaying newly committed sidecar log entries through the deterministic NOOSChain block path
• no-quorum probe: two voters are stopped, the remaining voter cannot falsely finalize a new block, quorum is restored, and commits resume
• callback auth rejection
• duplicate committed callback idempotency
• malformed callback rejection
• final chain/replay verification

Commands

npm run stability:go-raft-chaos

npm run stability:go-raft-chaos

Environment

• GO_RAFT_CHAOS_LOG_DIR: report directory, default ./go-raft-chaos-results.
• GO_RAFT_CHAOS_KEEP_DB: reserved for future dedicated schemas.
• GO_RAFT_CHAOS_DURATION_MS: reserved for future long-running loops.
• GO_RAFT_CHAOS_SEED: report metadata seed.

Limitations

This is crash-fault testing for the HashiCorp Raft sidecar bridge, not Byzantine testing. It does not certify production Raft behavior, autonomous membership orchestration, OpenBFT, CometBFT, or libp2p.