Release Notes - v0.83.0
v0.83.0
Section titled “v0.83.0”Title: Improve synchronization, storage, and operational controls
Version: v0.83.0
Release: v0.83.0
Dingo v0.83.0 improves synchronization, ledger state handling, service reliability, storage efficiency, and operational controls compared with v0.82.1.
Recommended Network Compatibility ⚠️
Section titled “Recommended Network Compatibility ⚠️”| Network | Compatible |
|---|---|
| mainnet | ⛔ |
| preprod-testnet | ✅ |
| preview-testnet | ✅ |
| musashi | ✅ |
Synchronization and recovery
Section titled “Synchronization and recovery”- Reduced misleading synchronization errors by recording expected validation snapshot changes during overlay rebuilds at DEBUG rather than ERROR; retry cooldown behavior remains unchanged.
- Corrected epoch boundary pool handling during Mithril imports. Dingo now preserves staged re-registrations, stores current and staged records with registration indexes, and applies effective or deferred VRF ownership to support refunds and header validation. Deployments must apply database migration v43.
- Added a Genesis mode Limit on Eagerness that stops selection and header delivery at the candidate intersection, pauses held peers, resolves density standoffs, reports eagerness gauges and
eagerness_held, and cancels waits during shutdown. - Enhanced Blockfrost query scalability by storing pagination for extended pools, retiring pools, DReps, and asset holders, propagating request cancellation, and persisting and backfilling
utxo_addressmetadata for consistent pagination. - Isolated header mismatch tracking per connection, so only the connection with the mismatch reaches persistent fork resynchronization thresholds.
Peer and service reliability
Section titled “Peer and service reliability”- Preserved peer headers that Dingo could not yet apply beyond the count limit within byte budgets of 24 MiB per connection for unapplied history and 64 MiB total, and walked retained history during fork recovery.
- Safeguarded chain recovery after failed batch restores by tracking changes to block and header state, while keeping explicit CLI flags reliable across node construction and helper paths.
- Limited requests to recycle peers that fall behind the local tip to one every 30 seconds across connections, while skipping or releasing timers for active or unschedulable requests.
- Refreshed a 30-second write deadline before every write to Kupo
/matchesstreams, allowing stalled clients to release read snapshots while the serverWriteTimeoutremains 0. - Permitted private archive destinations through the default-off
barkBlockDownloadAllowPrivateAddressessetting, configurable through config, YAML, environment variable, or flag, while retaining origin allowlisting, HTTPS, redirect validation, and disabled ambient proxies.
Storage and operations
Section titled “Storage and operations”- Introduced
leiosPersistenceRetentionSlotsandDINGO_LEIOS_PERSISTENCE_RETENTION_SLOTS. Positive values turn on hourly cleanup of persisted Leios endorser block manifests and bodies, while 0 retains all history. Startup validates the minimum value, and shutdown protects storage when cleanup workers cannot drain. - Adjusted Genesis peer admission to reject an advertised tip that is more than one Genesis window behind the wall clock, including at the ingress gate, preventing dead fork adoption.
- Configured Badger provider settings with
gcIntervalandgcDiscardRatio, which default to 5 minutes and 0.5. Dingo leaves value log cleanup disabled in load mode and now reports on disk size and reclaimed space metrics correctly. - Accelerated Linux builds and Docker image jobs by allowing them to start after normal Linux tests, while release creation still requires race tests.
Docs authored by Doc Holiday