Use ClickHouse when analytics outgrow PostgreSQL

Review by

2027-04-30

How to do it

ClickHouse passes the five tests. Its source opened on GitHub in June 2016, and 26.10.2 shipped on 24 October 2026. Two lines get long-term support, 26.3 and 26.8. The clickhouse/clickhouse-server image had about 330 million pulls on Docker Hub in October 2026. The Python client clickhouse-connect had about 8.0 million weekly PyPI downloads. It stores data by column, so a sum over a billion rows reads only the columns it needs. It speaks SQL, so agents and analysts already know it. Feed it from the main database as Read changes from the database log describes. Choose something else when the data fits on one machine and one person analyses it. DuckDB runs inside the process with no server, but its 1.0.0 came out on 3 June 2024, so record the exception.

Examples

SQL example

CREATE TABLE page_views
(
    viewed_at DateTime,
    shop_id UInt32,
    path String
)
ENGINE = MergeTree
ORDER BY (shop_id, viewed_at);

SELECT shop_id, count() AS views FROM page_views WHERE viewed_at >= now() - INTERVAL 30 DAY GROUP BY shop_id;

Python example

import clickhouse_connect

client = clickhouse_connect.get_client(host="clickhouse.internal")
views = client.query("SELECT shop_id, count() FROM page_views GROUP BY shop_id").result_rows

How to check

Each report has a time budget in its definition. A scheduled job runs every report and fails when one takes longer than its budget.

Related rules

• EVT-7 Read changes from the database log in Integration and events