- The ledgers are stored in a sqlite database in ledger file system.
- Ledgers are organized in shards inside primary folder.
- Raw inputs are saved as shards inside blob folder. No input files are created if neither inputs nor outputs are available.
- Both last primary shard hash and last blob shard hashes are subjected to consensus and necessary sync operations are performed if out of sync.
- Hpfs sync support setting single sync targets from both ends of the list. (Targets set from front are prioritized).
- Contract and ledger syncs and serving are performed independently.
- Creating two separate hpfs folders for contract and ledger file systems (contract_fs and ledger_fs).
- Added separate mounts for contract and ledger file systems.
- Added separate instances for contract serving and ledger serving.
- Added separate instances for contract syncing and ledger syncing.
- Modified cluster creating script to accompany folder name changes in contract folder.
* Refactoring hpfs code to a class so it can support multiple mounts.
* Refactoring hpfs serve into a class to support mulitiple mount serving.
* Refactoring hpfs sync into class to support multiple instances.
* Code improvements in hpfs_sync.
* Taking a sync target list for hpfs syncing target set.
* Terminology changes from state to hpfs.
* Syncing hpfs parents in priority order.
* Changing how the hash is calculate in hpfs response validate check.
* Increased stage time window.
* Introduced central state tracker.
* Used shared mutex for target state update.
* Refactored lcl and state sync checks in consensus.
* lcl/state sync loop wait time improvements.
* Restored signal handlers upon fork().
* Improved error handling of hpfs process kill scenario.
* Set pgid for forked processes for graceful sending of SIGINT.
* Implemented going observer mode, fixed genesis lcl retrieval issue and stage closing time.
* Fixed clearing all user output instead of consensed outputs
* Modified waiting time to improve performance.
* Fixed deadlock of waiting for insufficient peers because of recent changes.
* Removed initial waiting time for peer connections to start consensus.
* Remove considering stage when voting and considering lcl.
* Stage sync , lcl sync and candidate set changes.
* Implemented ledger close time resolution and fixed ledger retrieval issues.
* Code cleanup and added more comments on reliability changes.
* Added further comments and clenup.
Detect and request missing lcl history from another random node.
Sending lcl history response to a asked node.
Getting lcl history response and applying it.
Delete lcl that exceeds max ledger sequence.
