张炎泼

The following task to enable cluster-name check: - [ ] Cluster name has to be persisted in storage. - [ ] The implementation of `RaftNetwork` has to send cluster-name in...

Feature: databend-meta: watch for a ephemeral node.

1

**Summary** Similar to ephemeral nodes in etcd or zookeeper: When the connection that creates this node is closed, drop the node.

**Summary** Make databend-meta a distributed WAL to support distributed transactional write in databend. A distributed transaction can be done on top of a distributed WAL. Databend-meta can be the WAL...

## Changelog ### Change: Add `RaftStateMachine::save_snapshot()` method Previously, `RaftStateMachine::install_snapshot()` was responsible for two tasks: - Replacing the state machine with the given snapshot. - Persisting the snapshot to disk. This...

Improve: Ensure uniform membership config during config changes

4

## Changelog ##### Improve: Ensure uniform membership config during config changes This commit improves the handling of membership configuration changes to ensure that the system reliably transitions to a uniform...

RaftCore - mpsc -> StateMachine

11

> When using the Stream approach to eliminate memory allocation, the Stream requires a 'static lifetime as it must traverse from the RaftCore task to StateMachine task. This necessitates that...

Relax request ordering to lower the read-request latency

1

Using NoopIndex instead of CommitIndex, when handling `ensure_linearizable()`, and slightly relaxing the ordering for requests could lower the read latency: A read request does not need to wait for committed...

chore: add logging to query queue test

1

I hereby agree to the terms of the CLA available at: https://docs.databend.com/dev/policies/cla/ ## Summary ##### chore add logging to query queue test ## Tests - [x] Unit Test - [...

Replication progress corruption when node rejoins cluster within same term

2

## Summary We've identified a replication session isolation bug in raft-rs that can cause infinite retry loops when a node is removed and then re-added to the cluster within the...

Leader survival across restart may cause state machine inconsistency for external readers

1

### Description OpenRaft provides a feature that allows a leader to survive a service restart. When a node that was a leader before restart comes back quickly enough (before any...