GreptimeDB Flow Guide

Create GreptimeDB flow definition to do time window based aggregation on the fly while data ingested into GreptimeDB. This is how GreptimeDB's lightweight streaming engine works.

Flow is also known as:

Stream computing
Materialized View
Continuous aggregation

The workflow

To create GreptimeDB flow task, we should follow these phases:

Phase 1. Learn and understand GreptimeDB Flow

First, we read GreptimeDB flow definitions and how it works from the documentation.

There are doc pages available, use WebFetch to load and understand them:

Overview and quick example: https://docs.greptime.com/user-guide/flow-computation/overview/
Examples for continuous aggregation: https://docs.greptime.com/user-guide/flow-computation/continuous-aggregation/
SQL DDL for flow, sink table and related concepts: https://docs.greptime.com/user-guide/flow-computation/manage-flow/

Phase 2. Create flow tasks

We should try to understand:

time window the user wants to use
aggregation fields and group-by rules

Create the flow and sink table DDL. Note that we will always provide CREATE TABLE statement for sink table, together with the CREATE FLOW statement because it's required.

Phase 3. Verify the tasks

If greptimedb-mcp-server is available, we can use its execute_sql tools to execute the DDL. Try to generate some sample data, insert into source table, then verify the sink table via SQL SELECT statements. We can use MySQL style SELECT statements on GreptimeDB for most cases.

Reference