Peter Lee

@otoolep Thanks for providing the doc. Yes I original though is "if the strong read request located on follower then it would be rejected and need to retry to send...

@otoolep > The only thing is to be clear what it gets us over using a Read Consistency of `None` along with suitable values for `freshness` and `freshness_strict`. Maybe the...

@otoolep My original thought is to directly apply this change to the existing `Strong` mode. The context below is under Strong Mode Read: I think it would be a significant...

> BTW, some of the code you need already exists -- and I added a little more. PR #1923 now means cluster clients can fetch the commit index from any...

Ok, Got it. @otoolep Also, I found that, in [Store Layer](https://github.com/lalalalatt/rqlite/blob/a123aac03077555ba73e0ce2a2a12604a4fd96a2/store/store.go#L254), rqlite treat query command just like other types of command, they're both called `raft.Apply(req)`. https://github.com/rqlite/rqlite/blob/a123aac03077555ba73e0ce2a2a12604a4fd96a2/store/store.go#L1234-L1235 Although it provide a...

> can you write what I would add to the existing technical documentation on Read Consistency? In other words don't just tell me what it does, tell me why I,...

> In fact the Leader that the Follower contacts may think it's the Leader, but it's not possible for a node that think it's Leader to be 100% **sure** it's...

> I have no idea if rqlite's use of the Hashicorp Raft code uses Leader Leases. You'd need to check that I take a quick look and just FYI: rqlite...

- [ ] add test for indexed follower read - [ ] multi-node tests where all followers are up (query should work) - [ ] multi-node tests where a quorum...

@ivandika3 Could you take a look at whether this API change makes sense? Since rewrite is intended for existing keys, allowing a semantic like “don’t create if existed” can be...