Node initial synchronization (DEPRECATED) - hadescoincom/hds-core GitHub Wiki

-- DEPRECATED --

HDS Node initially operates in Sync mode, then eventually it switches to the Standard mode.

Initial state selection is performed according to the following logic:

  • If the current blockchain tip is non-empty, i.e. contains at least the genesis block Standard mode is selected.
  • If there is a Treasury block(s) defined - the Node is supposed to create the genesis block. It switches to the Standard mode, mines the genesis block(s), and eventually broadcasts it to the network.
  • Otherwise Sync mode is activated.

The goal of the Sync mode is to allow the fast-Sync via Macroblock whenever possible. It consists of three phases.

  1. Detection phase. Node analyzes its peers, and selects the most suitable Macroblock to download.
  2. Download phase. Node downloads the Macroblock (by portions), from one or several peers.
  3. Import phase. Once fully downloaded - the Node uses it to import, and then switches to the Standard mode.

Synchronization via Macroblock

The Macroblock is a "compressed" blockchain history, which includes the following:

  1. All the block headers, with PoW.
  2. All the transaction kernels.
  3. The unspent UTXO set.

In contrast to the original blocks the macroblock doesn't contain any info about spent UTXOs. Once downloaded, the Node verifies the following:

  1. All the headers are valid (sane, have valid PoW, form a valid blockchain)
  2. All the original kernels are included in the macroblock
    • This is done by verifying groups of kernels corresponding to a specific block w.r.t. kernel commitment in the appropriate header.
    • This proves that all the original transactions are included.
  3. The resulting state of the system is valid:
    • Unspent UTXOs with all the kernels form a valid transformation from the genesis to the asserted height.
    • The overall value equals what it should be according to the emission schedule.
    • Recent coinbase UTXOs are time-locked, according to the system rules.

Detection phase

For each peer with non-empty tip the Node requests the following:

  • Chainwork proof (to ensure the credibility of the reported tip)
  • Information about the most recently generated Macroblock (its State ID).

This continues until either enough such peers are examined, or the timeout expires (since the first suitable peer was found). The macroblock of the peer with the highest Chainwork is selected. If there is no macroblock to download (which is typically the case until blockchain height is low) - the Node switches to the Standard mode, otherwise the Download phase is activated.

Download phase

During this phase the Macroblock is downloaded by portions. Each time it gets a portion - the next portion is requested from the same peer. In case the peer goes down or has no macroblock for upload - another peer is selected.

In case of shutdown/restart - the Node will continue the download from where it stopped.

Import phase

The Node performs the import from the downloaded Macroblock, and then switches to the Standard mode.

Notes

In case something goes wrong during the Sync mode - there is no automatic retry/restart. User intervention is required. Whereas in typical scenarios problems are unlikely, possible abnormal cases are:

  • No Macroblock detected. Can happen if the peer with the highest (proven) chainwork reports than it has no Macroblock. In this case Node goes straight to the Standard mode to sync, which is slower, and may not succeed in case old original blocks are already deleted in all peers.

  • Macroblock download stuck: Node attempts to download a Macroblock which no more exists in any of its peers. Can happen if the download takes insanely long, and during this time all the peers generated many newer Macroblocks. Typically each node keeps several most recently-generated Macroblocks (7 by default), and eventually deletes older ones.

  • Macroblock import error (data corruped, inconsistent state). Most probably the result of DoS attack, or less probably - some bug or the storage corruption.

In case of an error it's possible to reset the Node state, and retry the synchronization. Or, alternatively, the most recent Macroblock may be downloaded manually, and given to the Node for explicit initialization