본문으로 건너뛰기

Prune Node Data

This page explains how to delete historical block states to reduce the storage requirement. Kaia offers two approaches for pruning block states:

  • Live Pruning: With the live pruning feature enabled, the block states beyond a certain retention period will be automatically deleted.
  • Batch Pruning: State Migration): The block states can be state-migrated, meaning the block states before a certain block number become available.

Understanding Pruning Impacts​

"Live pruning" continuously deletes old states, keeping disk size at a minimum. However, due to accompanying bookkeeping tasks, live pruning slightly slows down block sync speed. "Batch pruning", on the other hand, does not affect performance after migration is complete, but a migration session takes a few days and temporarily requires large free disk space to copy the states.

How to Perform Live Pruning​

To enable live pruning from the genesis block, use the --state.live-pruning flag when starting your node. If you're starting from a database where live pruning is already enabled, the flag is optional but recommended for clarity.

노트

You can control the retention period for live pruning using the --state.live-pruning-retention NNN flag (default: 172800 seconds, that is 48 hours). This flag determines how long historical block states are kept before being pruned.

정보

Databases with and without live pruning are incompatible. To run a node with live pruning, you must either start from the genesis block with the --state.live-pruning flag or start from a chaindata snapshot that already has live pruning enabled.

You cannot convert a non-live-pruning database to a live-pruning database, and vice versa. Here are some example log messages you might see:


# First time enabling live pruning with an empty database
INFO[08/27,14:09:01 +09] [41] Writing live pruning flag to database
# Live pruning enabled
INFO[08/27,14:09:01 +09] [41] Live pruning is enabled retention=172800
# Live pruning disabled
INFO[08/27,14:09:46 +09] [41] Live pruning is disabled because flag not stored in database
# Cannot turn on live pruning after the chain has advanced (head block num > 0)
Fatal: Error starting protocol stack: cannot enable live pruning after chain has advanced

How to Perform Batch Pruning​

Prerequisites​

  • Recommended to run on a machine with m6i.8xlarge (32 cores and 128GB memory) or higher specs.
  • The machine should have enough spare disk space (500GB or more).
  • The entire process takes approximately 7 days to complete:
    • Stage 1: Copy (migrate) the state to a new directory. The message "State migration is completed" appears.
    • Stage 2: Block sync continues on the new directory. The old directory will be deleted after this step.

Steps​

  1. Attach to the node via console:

ken attach --datadir /var/kend/data

  1. Use the admin namespace RPCs to control state migration:

// Start
> admin.startStateMigration()
null
// Check progress
> admin.stateMigrationStatus
// Abort
> admin.stopStateMigration()