Skip to main content
Version: 3.1.0 (Preview)

Choose A Service

Choose a service based on what you want to run, then check its installation requirements below. The example store currently includes ten services.

This is a preview. The list describes the tested example packages, not every mode supported by the upstream software.

Available Services​

ServiceWhat it is forSoftware in the example store
NginxWeb serving and reverse proxyingFrom the operating system's package repository
PostgreSQLRelational data storageFrom the operating system's repository; PostgreSQL 10 in the reference test
KyuubiA shared SQL entry point for Spark1.9.4
AirflowScheduling and tracking workflows3.3.2
CelebornShuffle storage for computing jobs0.7.0
DolphinSchedulerBuilding and scheduling data workflows3.1.9
TrinoSQL queries across data sources483
DorisAnalytical data storage and queries4.1.4
ElasticsearchSearch and indexing9.5.4
MinIOS3-compatible object storageRELEASE.2025-10-15T17-29-55Z

Nginx, PostgreSQL, Airflow, Elasticsearch, and MinIO use the generic environment and do not need a Hadoop cluster simply to appear in the store. Kyuubi, Celeborn, DolphinScheduler, Trino, and Doris use the example BIGTOP environment. The page checks which services fit your destination; they cannot all be combined into any one cluster.

The store also contains a small foundation package. Ambari selects it when needed; it is not another application you need to run.

What Do I Need Before Installing?​

ServiceInitial setupPrepare first
NginxOne managed instanceA reachable OS package repository and a free port
PostgreSQLOne managed instanceOS packages, a persistent data directory, local management access, and backups
KyuubiServer and client componentsJava 17, Spark 3.5/Scala 2.12, Hadoop clients, and ZooKeeper configuration
AirflowOne host, LocalExecutorPython 3.11, a dedicated external PostgreSQL 14-18 database, and administrator details
CelebornOne master and one or more workersJava 17, the matching release archive, storage directories, and free ports
DolphinSchedulerMaster, worker, API, and alert processes on one hostExternal PostgreSQL 14-18, ZooKeeper, Java 11 or 17, and administrator details
TrinoOne coordinator, optional workersJava 25 on every assigned host, the release archive, and data directories
DorisOne FE and one BE, on the same or separate hostsARM64 archive, Java 17, enough disk space, and explicit credentials
ElasticsearchOne node using HTTPS and authenticationLinux/ARM64 archive, the required OS settings, storage, and administrator credentials
MinIOOne server with a consoleThe specified source release, Go 1.24.8, access to build dependencies or a prepared cache, storage, and new root credentials

Airflow and DolphinScheduler need a newer PostgreSQL database than the PostgreSQL 10 instance in the example store. Installing the PostgreSQL card is not enough for them.

Doris needs substantial temporary disk space: its archive is about 4.35 GB and expands to about 6.8 GB before you store any data. Leave room for both, plus previous installations and application data.

A Few Things To Know About The First Installation​

Airflow and DolphinScheduler need their own database space. Prepare an empty, dedicated database/schema and the right account. The installer initializes an eligible empty database and sets up the administrator; it does not take over someone else's existing data or automatically upgrade an existing schema.

Kyuubi uses an official binary distribution. Prepare compatible Spark and Hadoop dependencies before starting SQL engines.

Trino needs Java 25 even though Ambari itself uses Java 17. Its sample TPCH catalog is useful for a first query; connections to Hive, Iceberg, or other production data sources need your own configuration.

MinIO is built from the specified source release. The bundle does not contain a ready-made MinIO binary. Prepare the build environment and dependencies before installation; distributors also need to check upstream AGPL-3.0 requirements.

How Do I Know It Works?​

After installation and startup, run the service check from its Ambari service page. Checks exercise basic service behavior, such as a Trino query, a Doris write/read, or a MinIO object upload/download and cleanup.

If the check fails, open its task details and correct the reported problem. A running process alone is not enough to show that a query or object request works.

A passed check confirms basic operation. It does not replace capacity testing, backup recovery testing, or a production security review.

What Is Not Included Yet?​

The first packages target single-server or small-cluster use. Airflow CeleryExecutor, distributed MinIO, Celeborn high availability, and Doris replication/cloud mode are not included. Editing a setting does not add the missing deployment support.

You can edit full configuration files for all ten services. Monitoring support varies; seeing host CPU and memory charts does not mean every service has its own dashboard.

For the exact package revisions used in these examples, see the source baseline.