destinations.yaml decides how rows land

What destinations.yaml is

destinations.yaml declares the systems a job is allowed to write to. Its root destinations mapping gives each target a local identifier; pipeline.to selects one.

It does not choose the data or transform it. Sources read, transformations reshape, and a destination receives the final rows with the writing policy declared here.

What it contains

• Identity — the identifier referenced by pipeline.to.
• Reachability — type selects a connector and connection supplies its settings.
• Writing policy — load names the table or file, the mode, optional upsert keys, and batch size.

The write policy belongs with the target

The same six rows can replace a reporting file, append to a history, or update matching database records. That choice belongs beside the target because it describes how that system accepts rows, not how the rows were produced.

Run the complete declaration

This is the destination used by the tutorial job:

version: "1.0"
destinations:
  dest_big_orders:
    type: csv
    connection:
      base_path: "examples/tutorial/01-first-job/out"
    load:
      table: big_orders.csv
      mode: replace

hdrctl run examples/tutorial/01-first-job -d examples/tutorial/01-first-job/destinations.yaml

  ✅ Pipeline completed successfully in 7.8s
  Rows read   : 30
  Rows written: 6

The relevant command syntax, copied from hdrctl run --help:

hdrctl run [PATH]
  -d, --destinations FILE   explicit path to destinations.yaml
  --dry-run                 validate configuration without writing data
  -v, -vv, -vvv             more detail, up to full logs

Close by

• pipeline.yaml — selecting this destination
• sources.yaml — where rows enter
• transformations.yaml — shaping rows before the write
• Lesson 1 — the complete job
• Reference — every destination key