This patches an edge case bug introduced in v3.2.0 which could be encountered when a cosigner has downtime. The leader could get stuck, causing signing to stop even with threshold cosigners online.
Updating
If coming from v3.2.0-v3.2.2, this can be applied as a rolling upgrade, which prevents missed blocks:
- Shut down a non-leader cosigner, upgrade version, and restart
- Repeat for remaining non-leader cosigners
- elect one of the upgraded cosigners as the leader with
horcrux elect N - Shut down the initial leader cosigner, upgrade version, and restart
- elect initial leader as leader again if desired
If coming from ≤v3.1.0:
- Shut down all cosigners and delete
~/.horcrux/raftdirectory - Update binary or docker image
- Start cosigner nodes back up
What's Changed
- Fix stuck leader when cosigner goes down by @agouin in #232
- Reverse where we start checking for a non-expired nonce, oldest -> newest by @misko9 in #234
- add unit tests for prune by @agouin in #235
- mitigate unexpected state by @agouin in #233
New Contributors
Full Changelog: v3.2.2...v3.2.3