跳转至

文章背景与核心概要

Multigres 发布了其首个公开开源里程碑版本 —— v0.1 Alpha，为 Postgres 带来了 Vitess 级别的水平扩展、高可用性和操作简便性。作为专为 Postgres 设计的可扩展操作系统，该初始版本引入了高级连接 pooling（连接池）、基于共识的自动故障转移以及 Kubernetes 运营商（Operator），为未来的分片（sharding）功能奠定了基础。（注：针对 Supabase 的专属版本即将推出）。

Multigres v0.1 Alpha: An Operating System for Postgres

Summary: Multigres has released its first public open-source milestone—v0.1 Alpha—bringing Vitess-grade horizontal scaling, high availability, and operational simplicity to Postgres. Designed as a scalable operating system for Postgres, this initial release introduces advanced connection pooling, automatic consensus-based failovers, and a Kubernetes operator, laying the groundwork for future sharding capabilities. (Note: A dedicated release for Supabase is coming soon).

什么是 Multigres？

Multigres 是一个用于 Postgres 的可扩展操作系统，能够整体管理数据库实例，并提供内置的： * 分片（即将推出的旗舰功能） * 连接池 * 自动故障转移 * 备份编排

大规模运行 Postgres 会带来显著的运营复杂性，包括管理只读副本、故障转移、连接限制和备份。Multigres 将这些繁琐工作作为一个统一的整体系统进行处理，使您能够在需要时将数据库水平扩展。

v0.1 Alpha 具体引入了： * 高级连接池 * 自动故障转移 * 用于部署的 Kubernetes 运营商（Operator）

What is Multigres?

Multigres is a scalable operating system for Postgres that holistically manages your database instances, providing built-in: * Sharding (upcoming flagship feature) * Connection pooling * Automatic failover * Backup orchestration

Running Postgres at scale introduces significant operational complexity, including the management of read replicas, failovers, connection limits, and backups. Multigres handles these chores as a single cohesive system, empowering you to scale your database horizontally when needed.

The v0.1 Alpha specifically introduces: * Advanced connection pooling * Automatic failovers * A Kubernetes operator for deployment

Multigres 运营商（Operator）

Kubernetes Multigres Operator 允许您在 Kubernetes 上原生部署和管理 Multigres 集群。

  • 前提条件： 要开始使用，您需要一个 Kubernetes 集群和一个配置好的备份存储位置（例如共享文件系统或云存储桶，如 AWS S3）。
  • 本地测试： 您也可以在 Kind 集群上本地运行 Multigres。
  • 可用性： 运行 Multigres 所需的所有容器镜像均已公开可用。

The Multigres Operator

The Kubernetes Multigres Operator allows you to deploy and manage Multigres clusters natively on Kubernetes.

  • Prerequisites: To get started, you need a Kubernetes cluster and a configured location for backups (such as a shared file system or a cloud storage bucket like AWS S3).
  • Local Testing: You can also run Multigres locally on a Kind cluster.
  • Availability: All necessary container images for running Multigres are publicly available.

高可用性

Multigres 将高可用性（HA）视为一个共识问题，在不丢失成功提交的前提下成功解决了脑裂（split-brain）场景。其协议构建于通用共识之上，提供了独特的灵活性：

  • 基于未经修改的 Postgres 复制： 使用标准 Postgres 复制满足基于共识系统的一切严格一致性要求。
  • 用户定义的持久性策略： 定义任意复杂的持久性策略，而不受限制性多数派法定人数（majority quorum）规则的约束。例如，您可以强制要求跨可用区（cross-zone）持久性，同时将备用节点部署在三个以上的可用区中。
  • 动态副本： 在集群运行时安全地向上和向下扩展副本，而不会损害性能或正确性。

高可用性推荐阅读

High Availability

Multigres treats High Availability (HA) as a consensus problem, successfully resolving split-brain scenarios without losing successful commits. Its protocol is built on generalized consensus, offering unique flexibility:

  • Built on Unmodified Postgres Replication: Satisfies all strict consistency requirements of a consensus-based system using standard Postgres replication.
  • User-Defined Durability Policies: Define arbitrarily complex durability policies without being constrained by restrictive majority quorum rules. For example, you can mandate cross-zone durability while deploying standbys across more than three zones.
  • Dynamic Replicas: Safely scale replicas up and down while the cluster is running without hurting performance or correctness.

连接池

Multigres 采用基于双服务架构的自定义连接池解决方案： 1. Multigateway： 接受客户端连接并路由查询。 2. Multipooler： 管理后端数据库连接。

与传统的单进程连接池相比，这种架构具有几个明显的优势：

  • 智能流量路由： 与 HA 系统无缝集成，将连接路由到当前的主要是节点（primary）。在故障转移期间，它可以暂缓请求，直到提升新的主节点，从而将错误降至最低。它还支持跨多个副本的读负载均衡。
  • 上下文感知连接池： 具有内置的解析器，能够理解请求的影响，从而无需选择僵化的连接池模式（如事务级或会话级）。有状态的请求（如事务）会被固定到专用连接，直到完成。
  • 按用户划分的连接池： 为每个用户维护单独的、公平共享的连接池，无需共享池或 SET ROLE 模拟。
  • 预编译语句合并（Prepared Statement Consolidation）： 跨网关对预编译语句进行去重，因此 Postgres 对给定语句只需进行一次解析、计划和缓存。

连接池推荐阅读

Connection Pooling

Multigres features a custom connection pooling solution built on a two-service architecture: 1. Multigateway: Accepts client connections and routes queries. 2. Multipooler: Manages backend database connections.

This architecture delivers several distinct advantages over traditional single-process poolers:

  • Intelligent Traffic Routing: Seamlessly integrates with the HA system to route connections to the current primary. During failovers, it can hold requests until a new primary is promoted, minimizing errors. It also supports read-load balancing across multiple replicas.
  • Context-Aware Pooling: Features a built-in parser that understands request effects, eliminating the need to choose rigid pooling modes (like transaction or session-level). Stateful requests (like transactions) are pinned to a dedicated connection until complete.
  • Per-User Pools: Maintains separate, fairly shared connection pools per user without shared pools or SET ROLE impersonation.
  • Prepared Statement Consolidation: Deduplicates prepared statements across gateways so Postgres only parses, plans, and caches a given statement once.

备份

Multigres 利用 pgBackRest 进行可靠的备份，主要从副本执行备份，以避免给主数据库带来负载。

  • 三种备份类型：
  • 完全备份（Full）： 在检查点复制整个数据目录。
  • 增量备份（Incremental）： 仅复制自上次备份以来更改的文件。
  • 差异备份（Differential）： 复制自上次完全备份以来的更改。
  • 按需与定时： CLI 允许您列出备份、触发手动备份以及启动恢复。通过集群规范实现的定时备份即将推出。
  • 自动引导（Automated Bootstrap）： 自动识别主节点，执行备份，并使用它来初始化其他副本，无需人工干预即可启动准备就绪的集群。

引导推荐阅读

Backups

Multigres leverages pgBackRest for reliable backups, executing them primarily from replicas to avoid putting load on the primary database.

  • Three Backup Types:
  • Full: Copies the entire data directory at a checkpoint.
  • Incremental: Copies only files changed since the previous backup.
  • Differential: Copies changes since the last full backup.
  • On-Demand and Scheduled: The CLI allows you to list backups, trigger manual backups, and initiate restores. Scheduled backups via the cluster spec are arriving soon.
  • Automated Bootstrap: Automatically identifies a primary, performs a backup, and uses it to initialize other replicas, bringing up a ready-to-run cluster without manual intervention.

“Alpha”意味着什么

版本 0.1 已经足够稳定以供实验和反馈，但尚未准备好用于生产环境的工作负载。请牢记以下几点：

  • 还有一些已知问题需要解决。
  • 分片未包含在此版本中（v0.1 是一个具有 HA 和连接池功能的单分片集群）。
  • 不保证未来版本能够向后兼容。
  • 自定义资源（CR）API 不稳定，在 v1.0 之前可能会发生更改。
  • 性能基准测试目前正在进行中，将在后续文章中发布。

What "Alpha" Means

Version 0.1 is stable enough for experimentation and feedback, but not yet ready for production workloads. Keep the following in mind:

  • There are known issues that still need to be addressed.
  • Sharding is not included in this release (v0.1 is a single-shard cluster with HA and pooling).
  • Future releases are not guaranteed to be backward compatible.
  • The Custom Resource (CR) API is unstable and subject to change before v1.0.
  • Performance benchmarks are currently in progress and will be published in a follow-up post.

立即体验 Multigres

  1. 部署您的第一个集群： 使用自定义资源搭建一个最小的 3 节点 HA 集群。
  2. 加入社区：
  3. 在 GitHub 上提交问题并请求新功能。
  4. 在 GitHub 讨论区中讨论功能并获取帮助。

Try Multigres Today

  1. Deploy your first cluster: Stand up a minimal 3-node HA cluster using a custom resource.
  2. Join the community:
  3. File issues and request features on GitHub.
  4. Discuss features and get help in GitHub Discussions.