ClickHouse

Managed ClickHouse on StackBlaze: a fast columnar OLAP database with single-node or replicated high-availability deployments and automated backups.

Creating a ClickHouse instance

From the StackBlaze dashboard, click New service → Database → ClickHouse. Choose a deployment type (Standard or High availability), an initial storage size, and a compute plan. StackBlaze provisions the instance on Kubernetes with persistent storage attached.

Once created, open a web service or worker, go to the Environment tab, and connect the database. StackBlaze injects CLICKHOUSE_URL as a secret environment variable so your app can reach the instance without any manual wiring.

Connecting

ClickHouse exposes an HTTP interface on port 8123 and a native TCP protocol on port 9000. The HTTP interface is the simplest way to run queries from application code or tools like curl. Services in the same project reach the instance over the private network using its internal hostname.

HTTP endpoint (internal)

http://user:password@[service-name]:8123/?database=dbname

Use the native port 9000 with clients that speak the native protocol, such as clickhouse-client, for lower overhead and streaming inserts.

Standard vs high availability

A Standard instance is a single ClickHouse server. It is a good fit for development and for analytics that fit comfortably on one node.

A High availability deployment uses sharding and replication. Replicated tables (using the ReplicatedMergeTree family of engines) are coordinated by ClickHouse Keeper, so each write is copied across replicas. Data survives the loss of a node, and shards spread data across servers so the cluster can scale beyond a single machine. You choose the deployment type when creating the instance.

	Standard	High availability
Topology	Single server	Sharded and replicated
Replication	None	ReplicatedMergeTree via ClickHouse Keeper
Survives node loss	No	Yes
Scales beyond one node	No	Yes (sharding)
Best for	Dev, single-node analytics	Production, durability, scale

Querying

The HTTP interface accepts SQL in the request body. Using the injected CLICKHOUSE_URL, you can run a query directly with curl.

Query over HTTP

curl "$CLICKHOUSE_URL" --data-binary "SELECT version()"

ClickHouse uses table engines to control how data is stored. The MergeTree engine is the standard choice for analytical tables.

Create, insert, select

CREATE TABLE events (
  id    UInt64,
  name  String,
  ts    DateTime
) ENGINE = MergeTree()
ORDER BY (ts, id);

INSERT INTO events VALUES (1, 'signup', now());

SELECT name, count() AS n
FROM events
GROUP BY name
ORDER BY n DESC;

Backups

StackBlaze takes automated, encrypted backups of your ClickHouse instance following the platform backup policy. Backups are encrypted before leaving the host and can be restored from the database dashboard. For schedules, restore steps, and how backups work across services, see Backups & recovery.

Tip

Define an explicit ORDER BY key on your MergeTree tables. It determines how data is sorted on disk and has a large effect on query performance for analytical workloads.

Under the hood

ClickHouse is a columnar OLAP database built for fast analytical queries over large datasets. StackBlaze runs it via a Kubernetes operator with persistent volumes for storage. In high availability mode, data is replicated across nodes using ClickHouse Keeper so the cluster survives the loss of an individual node.