Snaphots documentation

[debilausaure]

Hi, I'm trying to implement replication in my application using openraft. I've read carefully about the protocol, but I'm not familiar at all with the implementation side.

My storage contains sensitive data, and the idea is to serialize/deserialize data on the storage side before passing it to openraft, which will only process opaque data blobs.
Essentially, openraft would receive receive write data from the storage whenever something needs to be written. It will first pass it in opaque form through a write API call, which will be sent as a request to other nodes and set into the log as a new entry using the append method from the trait RaftLogStorage (I am aware this needs to be persisted on storage). Once the log is propagated and the entry is committed, the storage will update its actual data though the apply method of the RaftStateMachine trait. On that side, I think I've got a rough idea of what to implement.

What I can't quite wrap my head around is how to handle snapshots. In the examples, snapshots are simply a copy of the state machine. I can't do that since my data is too large to fit in memory, but the method build_snapshot from the SnapshotBuilder trait specifies that it should.

Am I at a dead end here or am I missing something ?

[drmingdrmer]

A snapshot does not have to contain all of the data. It can be just a manifest file containing pointers to data files on disk.

When transmitting a snapshot via RaftNetworkV2::full_snapshot(), this method should send the data files to the remote peer and finally send the manifest itself.

Handling large snapshots that cannot fit in memory is a typical use case, and there should be a working example for this scenario. Unfortunately, this work has been delayed.

[debilausaure]

Thank you for your answer