Skip to main content

Recovering a Node

When a node is stopped (e.g. due to an upgrade or network outage), it may miss some blocks and fall out of sync with the network. The pages below describe a few options for recovering and rejoining the tip of the chain.

Each option has tradeoffs:

  • soft reset is fast but only works if the node tip is close to the network tip. It skips execution of the intervening blocks, so no artifacts (logs, receipts, traces) will be produced locally for those blocks
  • hard reset is slow but works even if the node tip is far from the network tip. It skips everything before the snapshot, so no artifacts (logs, receipts, traces) will be produced locally for those blocks.
  • full replay is more expensive, but it ensures the local archive has no gaps that would have to be served by S3. Full replay may be desirable for RPC providers.
  • node migration allows validators to achieve high availability by promoting a synced full node to validator status. This is useful for planned or unplanned maintenance with minimal downtime.

Soft Reset

Utilizes statesync to catch up to the tip of the chain, skipping execution of blocks in between.

Hard Reset

Restores state from a snapshot before resyncing.

Full Replay

Fetches and replays all missing blocks serially so that all transactional artifacts are available.

Node Migration

Promotes a synced full node to a validator by migrating configuration files and keys with minimal downtime.