Nodes use Serial LevelDB databases to persist processed blocks, transactions, and so on.
The data can be removed or not, depending on the pruning flags that can be enabled or not in
The flags used to specify if a node should delete databases or not are
Older versions of the configuration only have one flag
CleanOldEpochsData. If set to false, then old databases won't be removed.
By default, validators only keep the last 4 epochs and delete older ones for freeing disk space.
The default databases directory is
<node-working-directory>/db and it's content should match the following structure:
/db └── <chain id> ├── Epoch_X │ └── Shard_X │ ├── BlockHeaders │ │ ├── 000001.log │ │ ├── CURRENT │ │ ├── LOCK │ │ ├── LOG │ │ └── MANIFEST-000000 │ ├── BootstrapData │ │ ├── 000001.log | ............. └── Static └── Shard_X ├── AccountsTrie │ └── MainDB │ ├── 000001.log .............
Nodes will fetch the state from an existing database if one is detected during the startup process. If it does not match the current network height, it will sync the rest of the data from the network, until fully synced.
Starting a node with existent databases
There are use-cases when a node can receive the entire database from other node that is fully synced in order to speed up the process.
In order to perform this, one has to copy the entire database directory to the new node. This is as simple as copying the
directory from one node to the other one.
The configuration files must be the same as the old node, except the BLS key which is independent of databases.
Two nodes in the same shard generate the same databases. These databases are interchangeable between them. However, starting
a node as observer and setting the
--destination-shard-as-observer so it will join a pre-set shard, requires that it's database
is from the same shard. So starting an observer in shard 1 with a database of a shard 0 node will result in ignoring the database
and network-only data fetch.
If the configuration and the database's shard are the same, then the node should have the full state from the database and start to sync with the network only remaining items. If, for instance, a node starts with a database of 255 epochs, and the current epoch is 256, then it will only sync from network the data from the missing epoch.