Deployment monitoring and recovery tooling

2026-10-20 09:49:41 -04:00
parent 9d61c177b7
commit 55f26d3fad
26 changed files with 476 additions and 8 deletions
@@ -109,6 +109,8 @@ This mines three real blocks using disposable identities. It changes the testnet

 ## Network and safety limits

+[Deployment and recovery preparation](docs/operations-recovery.md) includes explicit trusted-proxy configuration, opt-in loopback-only diagnostics and `ops:check`, plus stopped-node paired `node:backup` create/verify/new-directory restore tooling. Backups contain sensitive market data and require external encryption/access control. Tools do not configure TLS/alerts or authorize production launch/restore.
+
 [Stage-1 node resource controls](docs/node-resources.md) bound global RPC admission, response bytes and canonical page sizes. The UI reuses paged history; legacy full-state responses stop beyond 128 blocks. A consent-gated loopback-only smoke-load script is available as `test:rpc-load`. This does not resolve long-chain/DAG storage limits or establish production capacity.

 Nodes use only explicitly configured peers; there is no discovery service or public bootstrap network. The peer, storage, and HTTP configuration is documented in [docs/testnet-consensus.md](docs/testnet-consensus.md). The implementation is experimental: it has no consensus transaction fees or established economic security (market fees are off-chain simulations), and remains vulnerable to resource exhaustion, isolation/eclipsing, censorship, majority-hashpower attacks, and reorganizations. The ledger now uses transactional SQLite/WAL storage and sparse per-block NFT-state deltas, but nodes still retain DAG block records in memory, rebuild state roots over the full NFT set, and revalidate all proofs at startup. It is not yet an archival-scale production design. HTTPS, container hardening, and passing tests do not constitute a security audit.