Skip to content

Implement saving and reading persistent states with parts in S3 #1025

Description

@SmaGMan

Scope

  • Extend the S3 persistent-state loader to support persistent states with parts, not only single-file persistents.
  • First attempt to fetch meta.json from S3; if present, read part_split_depth, derive shard part prefixes, and build the FoundState response with parts. Individual parts should be fetched via separate requests.
  • Extend request arguments in the same way as BlockchainRpcService / BlockchainRpcClient.
  • When saving a persistent state to S3, use PersistentState.CachedState.parts_info to discover parts; each part includes a prefix to load its chunks via read_state_chunk.
  • Save meta.json into S3 so that the persistent state with parts can be discovered and downloaded correctly.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

C-coreCore moduleC-rpcRPC subscriber moduleC-storageStorage moduleP-mediumMedium priorityS-featureSeverity: feature. This is adding a new feature.

Type

No type

Projects

No projects

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions