Release Notes - v0.94.0
Este contenido aún no está disponible en español.
v0.94.0
Section titled “v0.94.0”Title: Improve ledger safety, synchronization, and node operations
Version: v0.94.0
Release: v0.94.0
Comparison: v0.93.0…v0.94.0
Dingo v0.94.0 improves remote key management, peer synchronization, ledger state safety, database performance, observability, and recovery behavior. It also updates protocol capabilities and strengthens operational safeguards across node services.
Recommended Network Compatibility ⚠️
Section titled “Recommended Network Compatibility ⚠️”| Network | Compatible |
|---|---|
| mainnet | ⛔ |
| preprod-testnet | ✅ |
| preview-testnet | ✅ |
| musashi | ✅ |
Remote access and synchronization
Section titled “Remote access and synchronization”- Added ordered failover across remote KES agents so node forging can continue when one remote service is unavailable.
- Named remote KES agent profiles so deployments can select a configured profile without changing the failover list.
- Strengthened historical peer validation so the node rejects older peer data when its block relationships are inconsistent.
- Separated catch-up pauses from peers that cannot provide enough chain work so chain selection can continue while catch-up is paused.
- Renamed the CBOR conformance report field to
dataset_commitso generated reports use the expected field name. - Retained rate-limited transaction offers until the retry budget becomes available, so the node fetches each offered transaction after a retry instead of dropping it.
- Served only ledger-applied blocks through ChainSync and BlockFetch, withholding unapplied blocks and rejecting invalid ranges without returning blocks.
Protocol, ledger, and validation
Section titled “Protocol, ledger, and validation”- Stabilized credential watcher validation by replacing watched files as a single operation, preventing partial credentials during validation tests.
- Expanded protocol capabilities through gouroboros v0.220.0, including DELEG support, corrected Dijkstra fee handling, and IPv6 relay fixes.
- Preserved imported Mithril outputs through rollback and safe truncation, keeping needed outputs available after cleanup.
- Batched Mithril immutable retirement to process several chunks together during cleanup.
- Expanded forge, reward, and halted pipeline health test coverage to verify these workflows under boundary conditions.
- Fixed governance vote history to record deletion markers, so at-slot and paginated reads omit votes during their deleted windows and rollback restores the correct state.
- Corrected certificate ordering so a Leios closure appears before its body, allowing readers to select the applicable certificate consistently.
- Maintained Byron boundary geometry for targets before the Shelley transition, preventing slot-order errors that can stall synchronization.
- Protected credited reward outputs when a late retry precompute runs after the system credits the reward round, so the retry cannot clear credited balances.
- Retained imported Conway committee state during Mithril startup, so startup does not reseat removed members while it still validates configuration.
- Restored rollback undo events for transactions in certified Leios endorser blocks, so consumers remove those transaction effects when the certifying block rolls back.
- Improved sparse empty epoch transitions and reward boundaries, keeping nonce state aligned and accepting the first Shelley block at the correct Byron transition.
- Aligned Leios validation retries with the post-commit tip signal, preventing validation from running before the ledger applies the block.
- Recorded removed genesis committee members during Mithril imports, keeping their authority absent across restarts and rollbacks.
Chain selection, recovery, and database performance
Section titled “Chain selection, recovery, and database performance”- Protected chainsync resynchronization from races by checking the same chain point before restoring peer history.
- Migrated SQL UTxO identity and spend tracking to transaction IDs for consistent transaction output ownership and lookup.
- Optimized spent-output lookup by indexing only UTxO rows that contain a deletion slot.
- Rejected pre-Conway ledger-state imports instead of accepting state from an unsupported era.
- Enforced slot order and point-based Mithril trust so imported chain state follows the correct position and trust boundary.
- Covered reward updates for epochs with no block to verify the empty-epoch transition path.
- Delayed and withheld chainsync-demoted peers so peer selection does not immediately reuse a peer that lost chainsync status.
- Prevented false rollback and divergence alarms after history expires by using retained local chain metadata to classify rollback points.
- Recovered from withheld header ancestry with a bounded re-intersection, allowing linkable descendants to reach the ledger without refreshing stalled peer activity.
- Streamlined local ChainSync recovery by re-intersecting the active connection in place, preserving peer state and avoiding unnecessary reconnects.
- Removed the redundant
idx_script_hashindex fromscript.hashthrough upgrade migration v65 across SQLite, PostgreSQL, and MySQL; hash lookups continue using the unique constraint index (uni_script_hashor SQLite’s unique autoindex), reducing duplicate index maintenance during inserts. This upgrade migration does not change external API or RPC behavior.
Operations, observability, and recovery
Section titled “Operations, observability, and recovery”- Resolved rebuilt reward owners at the snapshot window so reward ownership uses the correct historical state.
- Enabled Genesis ChainSync Jumping by default to accelerate initial synchronization.
- Introduced opt-in
pprofprofiling and trace regions for investigating mutex and block performance. - Pruned uncertified Leios endorser blocks behind the immutable tip to limit stale pending data.
- Secured Unix socket permissions to limit local service access.
- Reported the electing VRF key in pool distributions so election results identify the key used.
- Removed the duplicate datum index to reduce redundant database indexing.
- Published recovery rollback events so downstream consumers can respond to rollback recovery.
- Lowered routine node-to-client connection setup failures to informational logging, reducing error noise when clients close before the handshake completes.
- Enabled multi-pool forging on one node through
blockProducerPools, with separate pool operations and coordinated slot selection. - Evicted confirmed transactions from the mempool on every chain update, including while a rebuild retry waits to run.
- Clarified effective configuration logging and added ChainSync and pruning metrics for startup, rollback, deferred verification, and cleanup monitoring.
- Corrected Antithesis container directory ownership so the Unix socket listener can start successfully.
- Bounded Blockfrost epoch stake paging and made busy or slow requests return retryable responses with
Retry-After. - Delayed chain position gauge registration until the ledger tip loads, preventing misleading zero values and alert spikes during startup.
- Retired extracted Mithril chunks after committed batches and stored resume markers, reducing disk usage and avoiding downloads of deleted chunks after interruption.
- Enabled unrestricted API defaults with positive opt-in limits, allowing operators to choose admission and pagination caps explicitly.
- Improved ledger reader test reliability by canceling and awaiting reader goroutines before test cleanup; this change strengthens validation coverage only.
- Strengthened Leios retry test semantics by requiring a post-commit retry; this change provides deterministic validation coverage only.
Docs authored by Doc Holiday