raft-dev

https://github.com/raft/raft.github.io/pull/216 https://github.com/glycerine/rpc25519/tree/master/

I've gotta say I wish someone had pointed me to the MongoRaftReconfig algorithm paper earlier.

Hi Jesse, I was thinking of the simplest data model, in which there is (a) a state machine, and (b)

Thanks guys! On Tuesday, September 2, 2025 at 2:54:26 PM UTC+1 je...@mongodb.com wrote: My colleague

Hi AJ, On bootstrapping, Section 4.4, page 46 of the dissertation says, "Instead, we recommend

This has started happening again about two days ago: frequent downloads of the user study raft.mp4

Hi, we don't need to replace the vote request handling from ``` if req.term >= self.term ```

Some Raft newcomer questions: The Raft dissertation and talks mention that the commit index

You are right. This algo can safely change the config. But it brings in other issues: when a node

Yes. Persisting the commit index offers clear benefits: upon startup, the server can immediately

I think from Raft fault tolerance assumption, N = 2f + 1 nodes is required to tolerate upto f crash

Turns out I'm using SSDs already, not EBS. On Saturday, January 18, 2025 at 7:42:37 AM UTC-5 A.

Something like this (see "In Search of an Understandable Consensus Algorithm" §8 for

Hi Diego. Thanks for replying. So we can say that the leaders of T+1, T+2, ..., U-1 (if those terms

Hi, Has anyone tried to run the raft cluster formation across two or more k8s clusters? There are two

You should have just one node at first and add the other two one by one later to the cluster. On

Don't call Bootstrap if there is pre-existing state. Just called NewRaft, and pass it the

Thanks! On Tuesday, October 29, 2024 at 5:17:29 PM UTC-7 Philip O'Toole wrote: You just call

On Tue, Oct 15, 2024 at 5:10 PM Diego Ongaro <onga...@gmail.com> wrote: You've linked to

Reading committed values (even stale) may sometimes be ok but there may be an issue if the client who

That makes sense. Pipelining is a significant benefit! Thanks, Andy! M. On Tuesday, October 1, 2024

1. If the leader comes back online before the election timeout is exhausted on the followers then it

The state machine's idempotency is critical here since RAFT cannot guarantee exactly-once command

Hi saurabh ratna, Since that sounds very specific to the braft implementation, you might want to ask

For a crashed node, we can assign it a new UUID and use reconfiguration to remove the one with old

Actually, I have only a purpose is performance analysis for raft and PoW consensus protocols. As a

Yup my bad, thank you for the clarification! Will it be safe to assume that the persistent state

Sorry if what I said was ambiguous... The majority *may* include the leader, but doesn't have to,

On Mon, Jun 3, 2024 at 2:58 PM H Wang <howard.w...@gmail.com> wrote: a fresh new S1,

Replicated logs should use the original term number. It is not safe to change them to the new term