How to Install ClickHouse on Debian

ClickHouse is a high-performance, column-oriented database built for analytics, logs, metrics, and real-time queries. Debian is one of the best environments for running ClickHouse because of its stability and clean package management.

Step 1: Update System Packages

Start with a clean and updated system.

sudo apt update && sudo apt upgrade -y

Step 2: Install Required Dependencies

ClickHouse uses HTTPS repositories and GPG verification.

sudo apt install -y apt-transport-https ca-certificates curl gnupg

Step 3: Add the ClickHouse GPG Key

Import the official signing key:

curl -fsSL https://packages.clickhouse.com/CLICKHOUSE-KEY.GPG | sudo gpg --dearmor -o /usr/share/keyrings/clickhouse.gpg

Step 4: Add the ClickHouse Repository

Create the repository source file:

echo "deb [signed-by=/usr/share/keyrings/clickhouse.gpg] https://packages.clickhouse.com/deb stable main" | sudo tee /etc/apt/sources.list.d/clickhouse.list

Update package lists:

sudo apt update

Step 5: Install ClickHouse Server and Client

Install both the server and CLI tools:

sudo apt install -y clickhouse-server clickhouse-client

Step 6: Start and Enable ClickHouse Service

Enable ClickHouse at boot and start it immediately:

sudo systemctl enable --now clickhouse-server

Check service status:

systemctl status clickhouse-server

You should see active (running).

Step 7: Connect to ClickHouse

Access the ClickHouse client:

clickhouse-client

Test the installation:

SELECT version();

If a version is returned, ClickHouse is working correctly.

Step 8: Allow Remote Connections (Optional)

By default, ClickHouse only listens on localhost.

Edit the config file:

sudo nano /etc/clickhouse-server/config.xml

Find and modify:

<listen_host>::</listen_host>

Restart ClickHouse:

sudo systemctl restart clickhouse-server

If UFW is enabled, open the native TCP port:

sudo ufw allow 9000/tcp

ClickHouse is not a replacement for MySQL or PostgreSQL. Use it where it shines: analytics, time-series data, logs, metrics, and massive read-heavy workloads. If you try to use it like a transactional database, you’re using the wrong tool.