Principle:Apache Druid Spec Review and Task Submission
| Knowledge Sources | |
|---|---|
| Domains | Data_Ingestion, Task_Management |
| Last Updated | 2026-02-10 00:00 GMT |
Overview
A task submission principle that validates, reviews, and submits a complete ingestion specification to the Druid Overlord for execution.
Description
Spec Review and Task Submission is the final step of the batch ingestion wizard. The user reviews the complete ingestion specification as a JSON document in an inline editor, makes any manual adjustments, and submits it to the Druid Overlord's task API.
The submission triggers an indexing task that:
- Reads data from the configured input source
- Applies parsing, timestamps, transforms, and filters
- Creates segments according to the schema and partitioning configuration
- Publishes segments to deep storage and registers them in the Druid metadata store
The Overlord returns a task ID that can be used to monitor progress, view logs, and check completion status.
Usage
Use this principle as the terminal step of any batch data ingestion workflow. All prior configuration steps must be completed to produce a valid ingestion spec. The submission is irreversible — once submitted, the task begins consuming source data.
Theoretical Basis
Task submission follows a spec validation and dispatch pattern:
IngestionSpec (JSON) → POST /druid/indexer/v1/task
Response → { task: string } // Task ID
Task lifecycle:
PENDING → RUNNING → SUCCESS | FAILED
The Overlord validates the spec, allocates workers, and begins execution.
Failed tasks can be retried by resubmitting the same spec.