Does a RaftError always indicate that a node is broken and must shutdown?

[semirix]

When a RaftError is returned from something like client_write(request) I assume RaftError::ShuttingDown means that the node is in the process of shutting down and that the error that caused it already happened. If I get any of the other variants in RaftError like RaftError::SnapshotMismatch should I assume that the entire node is in a broken state and must be shutdown?

[drmingdrmer]

I assume you are using v0.6.4, in this version RaftError::SnapshotMismatch is only used internally thus an application should have no chance of receiving such an error.

I'd recommend using the latest main branch in which the errors are better defined:
client_write() will return an ClientWriteError:

pub enum ClientWriteError<NID: NodeId> {
    #[error(transparent)]
    ForwardToLeader(#[from] ForwardToLeader<NID>),

    /// When writing a change-membership entry.
    #[error(transparent)]
    ChangeMembershipError(#[from] ChangeMembershipError<NID>),

    #[error(transparent)]
    Fatal(#[from] Fatal<NID>),
}

ClientWriteError is much more obvious for an application to deal with.

[semirix]

Is there a particular commit I should pin my dependencies to until the next version release?

[drmingdrmer]

No.

Pinning to the latest main would be OK.
The latest main branch has several changes from 0.6.4, the example-raft-kv dir would help.

In the next several weeks there will be some internal refactoring commits that won't affect the API.
After that, there will be several API-changing commits, in order to provide more efficient storage access.

Please be aware of commits start with Change:, which brings in breaking changes.