Introduction

Cobble Flink is unified storage for Apache Flink: write processing results, read stored data, and manage streaming state through the same Cobble core.

Watch the Introduction

Use Cases

Balance storage cost and state performance

Separate storage from compute, starting with Flink 1.17. Choose hybrid storage or local caching to balance capacity, cost, and performance.

Cobble core manages local and remote storage with hybrid or local-cache policies.

Explore the state backend · See benchmarks

Offload compaction from processing jobs and share CPU capacity across their peaks.

Illustrative remote compaction sizing: six provisioned CPUs become four with staggered peaks.

Configure remote compaction

Write once, reuse across jobs

Let another job scan stored results or look up a key for enrichment. Supported keyed state is reusable too.

A Flink writer publishes Cobble snapshots for other jobs to scan or look up by key.

Write with the sink · Scan and look up data

See what changed in your state

Look up state and sink records, inspect business fields, and track changes across snapshots to debug your job.

Web Monitor illustration showing a customer record changing between completed checkpoints.

Explore the Web Monitor

Documentation Structure

Chapter	Description
Getting Started	Install Cobble Flink and choose a version
State Backend	Store Flink managed state in Cobble
Catalog	Create and use native Cobble SQL tables
Source	Read Cobble data from Flink SQL
Sink	Write Flink SQL results to Cobble
Web Monitor	Inspect checkpoints and tables
Metrics	Monitor state backend and connector activity