This is a quality-of-life patch release for operators.
Important
Operators using tempo download to start nodes from snapshots please upgrade
Update Priority
| User Class | Priority |
|---|---|
| Validators | Recommended |
| RPC Nodes | Recommended |
What's Changed
UX Improvements
- Fixes
tempo download --force#7256:
tempo download --forcenow deletes the entire consensus directory. - Fixes
tempo downloadnot fetching required consensus data #7154: operators no longer need to providetempo download --skip-consensus=falseto download required consensus data. This now is the default. - Fixes
tempo downloadrequiring explicit--datadirand--manifest-urlarguments #7264: operators no longer need to provide these arguments (unless desired). With v1.13.2 the node uses the Tempo snapshot API to determine which manifest to download and correctly resolves the datadir.
Bug Fixes:
- Fix rare race condition leaving consensus metadata invalid #7203: never observed in real deployments but identified as a potential issue that would require operator intervention.
- Fresh or stale nodes can participate in consensus faster #6950: validator nodes that have been offline and are restarted with a fresh snapshot, or nodes that are freshly spun up, will now attempt to reconstruct their signing share by reading their last known epoch. If successful, they will participate in consensus immediately instead of waiting for a full cycle.
- Fix nodes refusing to start up because they cannot find boundary headers #7244: in certain scenarios nodes running with
tempo node --minimaltried to read already pruned boundary blocks and shut down even though boundary headers were available and contained the desired data. - Noisily rejects startup on invalid data #7255: due to manual intervention a node could be brought into a state where its available finalization certificates were older than persisted node metadata. This state is now detected early and a node shuts down with a clear error message.
Full Changelog: v1.13.1...v1.13.2