Compute-Storage Decoupled Operations Guide

The compute-storage decoupled architecture separates compute nodes from remote shared storage, enabling independent elastic scaling of compute and storage. This guide helps you configure and operate a compute-storage decoupled cluster.

Cluster Operations

Compute Group Operations

Group one or more BE nodes into a compute group. Learn how to create, scale, and manage compute groups in compute-storage decoupled mode.

Managing Storage Vault

Configure the remote shared storage (Storage Vault) used in compute-storage decoupled mode. Supports multiple vault configurations and per-table storage location assignment.

Data Recycling

Learn how Doris safely and efficiently reclaims storage space occupied by deleted data in the compute-storage decoupled architecture, and how to manage the data lifecycle.

Upgrade

A step-by-step guide for upgrading a Doris cluster in compute-storage decoupled mode using rolling upgrades, minimizing impact on upstream workloads.

File Cache

File Cache Overview

Accelerate remote storage access with local disk caching. Covers the multi-queue LRU strategy, cache warm-up for new compute groups, and configuration details.

File Cache Internals

A deep dive into the file cache implementation: 1 MB-aligned slicing, prefetch mechanisms, and local filesystem block file management.

Read-Write Separation

Read-Write Separation

Support cross-availability-zone high availability and read-write separation architectures through the File Cache active incremental warm-up mechanism, reducing query latency and accelerating failover response.

Cache Optimization Best Practices for Read-Write Separation

In a multi-compute-group read-write separation deployment, optimize cache hit rates to prevent read-only compute groups from frequently pulling data from remote object storage, improving query performance.