The roadmap of the next sekas version

[w41ter]

This version aims to provide a basic fault-tolerant, scalable transactional key-value store.

Before releasing the new version, we need to solve below issues:

Features:

• [x] #42
• [x] Support CAS operation
• [ ] Support multi-version snapshot
• [x] #56
• [ ] Support set TTL for keys
• [x] Support watch API
• [x] Define the API of the Rust SDK
• [ ] Add a proxy to service the Rust SDK
• [x] Add a cli to control the cluster and fetch/write data
• [ ] Support shard automatic split/merge (for txn col, support prefix split)
• [ ] Support multi-version watermark and recycle tombstone records
• [ ] Support recycle TXN state table records
• [ ] Support a file service discovery to supply nodes
• [ ] Add cluster_id for all RPC requests, and validate it before execution.

Improvement:

• [x] Add task group to manage async task lifecycle
• [ ] Stop all serving replicas when dropping node, then pass test node::tests::create_then_bootstrap_replica.
• [ ] Migrate the code of raft_engine to a sub-crate for easier maintenance.
• [ ] Migrate the code of raft-rs to a sub-crate for easier maintenance.
• [x] Rename shard migration to shard move and refactor it
• [ ] Change the Master location query API from push to pull.
• [ ] Add cluster ID in the configuration file
• [x] Move the ID definitions from sekas-server/constants.rs to sekas-schema
• [ ] Reduce the test running time
• [ ] Support scan keys without returning the value set
• [ ] Use Error to replace AppError

Testing:

• [x] Add unit tests for eval_scan.rs
• [ ] Add snapshot-related unit tests for the raft_group mod
• [ ] Simplify the testing framework and ensure the stability of integration testing

Some things we plan to do in future releases:

• Backup/restore
• Data ingestion
• A storage engine
• A RPC framework, compatible with bRPC, gRPC
• A runtime to schedule tasks in a separate task group.