Cosmos Snapshot vs State Sync: How to Choose the Right Bootstrap Method
Compare Cosmos snapshots and state sync for node bootstrap, historical data, validator safety, and recovery.
A new Cosmos SDK node does not always need to replay the chain from genesis. Depending on its role, it can start from a data snapshot, use state sync, or replay blocks from an earlier point.
The right choice depends on what the node must do after it catches up. A dedicated RPC node serving current balances has different requirements from an indexer that needs historical blocks. A validator also needs a safe signing procedure that goes beyond restoring the database.
Snapshot vs state sync: the practical difference
A data snapshot is usually a copy of another node's data directory. You download the database, restore it on your server, and let the node catch up from the snapshot height.
State sync downloads application state from peers and verifies the restored state against a trusted header. The node then catches up from that point.
Both can reduce bootstrap time, but they make different assumptions:
- Starting point: A data snapshot copies an existing node's data. State sync obtains recent application state from peers.
- Trust: A snapshot relies on trust in its source and data. State sync relies on a trusted header and light-client verification.
- Historical data: Snapshot coverage depends on its contents. State sync does not reconstruct all earlier blocks or historical state.
- Compatibility: Snapshots depend on the database format and application version. State sync requires a compatible application and peers.
- Typical use: A snapshot can restore a node with known data coverage; state sync can bring a fresh node to recent state.
Neither method automatically creates an archive node. Historical block storage, historical application state, and transaction indexes are separate requirements.
DTEAM resources for Celestia mainnet
If you are preparing a Celestia consensus node, DTEAM provides a state sync page and a pruned snapshot. Choose the method based on the data your node needs, and check the network and software requirements before starting.
When a snapshot makes sense
A snapshot can be a good fit when you need to restore a node with a specific data set, or when the network does not have reliable state-sync peers.
Before downloading, check:
- Chain ID and snapshot height.
- Application version and database backend.
- Which blocks, application states, and indexes are included.
- Compressed and uncompressed sizes.
- Whether the archive contains configuration or sensitive files.
- Whether the publisher provides a checksum or signature.
A checksum helps detect a damaged download. It does not independently prove that a snapshot is trustworthy if the checksum and archive both come from the same compromised source.
If the provider calls a snapshot "archive," ask what that means in practice. A node may retain old blocks without retaining historical application state. Test the actual queries you need at the oldest required heights.
For more context, see Full Node vs Archive Node.
When state sync makes sense
State sync is useful when a node needs recent state and does not need to reconstruct the entire chain history.
A typical setup requires:
- A compatible node binary.
- Peers that serve application snapshots.
- RPC endpoints used by the light client.
- A trusted block height and hash.
- A trusting period appropriate for the network.
The trust height and hash should come from sources you trust. Listing the same RPC server twice does not create independent sources.
The following is a configuration example, not a production-ready configuration:
[statesync]
enable = true
rpc_servers = "https://rpc-a.example.com:443,https://rpc-b.example.com:443"
trust_height = 1234567
trust_hash = "REPLACE_WITH_VERIFIED_BLOCK_HASH"
trust_period = "168h"
Replace all example values and follow the instructions for the specific network and installed software version. The example trusting period is not suitable for every chain.
State sync can finish successfully while the node still lacks older blocks or historical state. If your application needs those records, test them separately.
Celestia's consensus node guide provides one network-specific example of state sync and its trust requirements.
Validator recovery requires extra care
A validator is different from an ordinary RPC node because it signs consensus messages. Restoring a database without preserving signing safety can create a risk of double signing.
Do not run a generic reset-and-restore script against an active validator.
A safer approach is to restore and verify a separate, non-signing node first. Keep production signing keys off that machine until the migration procedure is ready. Before moving signing responsibility, disable the old signing path and follow the documentation for the exact signer in use, including its authoritative signing state.
A local signer and a remote signer can store safety information differently. Copying one file is not a complete validator failover plan.
Why bootstrap attempts fail
Several problems appear repeatedly during node recovery:
No compatible peers. A working RPC endpoint does not guarantee that state-sync snapshot peers are available.
Version mismatch. A data snapshot may depend on a particular application version or database backend.
Insufficient storage. The extracted database, temporary files, catch-up, and future growth all need space. The compressed archive size is only part of the calculation.
Unexpected pruning. A node can start from historical data and later remove it under its configured retention policy. Check block retention, state pruning, and transaction indexing separately.
Unverified restore. A node reporting that it is caught up does not prove that it serves every query the application needs.
For production, measure the full recovery process: provisioning, download, extraction, catch-up, query checks, and return to service.
Dedicated Cosmos nodes with a recovery plan
If your application depends on Cosmos RPC or historical data, bootstrap is only one part of the operating plan. You also need monitoring, capacity planning, upgrades, and a tested path to recovery.
DTEAM can provide dedicated nodes for Cosmos networks, including networks not currently listed on our website. Tell us which chain you need, the interfaces your application uses, and how far back its queries must go. We can then discuss the node configuration and recovery approach.