Skip to content

Discuss chdb-durable checkpoints for HyperbyteDB recovery #111

Description

@ShawnChen-Sirius

I noticed that HyperbyteDB already has serious recovery-related design: WAL, metadata, clustering, and derived chDB state. That makes this especially interesting, because chdb-durable may be able to complement the current architecture.

chdb-durable is a durable mode for embedded chDB: it stores chDB state in object storage or a local durable namespace, restores from checkpoints plus WAL replay, and uses a single-writer lease.

Potential value for HyperbyteDB:

  1. Restore chDB-derived state faster during cold recovery.
  2. Reduce the amount of existing WAL that needs to be replayed into chDB.
  3. Help new nodes bootstrap from durable chDB checkpoints.
  4. Complement existing archival/replication design without replacing it.

Today chdb-durable is mainly exposed from Python. If this direction sounds useful, the chDB side can help contribute a Rust durable API or sidecar and discuss the correct boundary with HyperbyteDB's WAL, schema/Raft, manifest, and recovery flow.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions