Configuration overrides

Coded addPipeline({ ... }) remains the source of truth for structure (source entity/service, target, adapters, query closures, view mapping). Tunable operational knobs can be overridden at runtime through the management API and the Pipeline Console. Overrides are persisted in the tracker and re-applied after every process restart — because pipelines always re-register from code on boot, the engine loads the stored JSON delta in _ensureTracker and layers it onto the fresh coded baseline.

coded addPipeline(config)
        │
        ▼
  baseConfig  (normalized, in-memory + tracker.baseConfig)
        │
        │  deep-merge with
        ▼
  overrides   (tracker.overrides JSON)
        │
        ▼
  effective = Pipeline.config  (what runs actually use)

Overridable fields

Path	Effect
enabled	Pause scheduled ticks (false); manual start / execute still run
mode	Default run mode (delta / full / partial-refresh)
schedule	Interval (ms / time string), cron, and engine (spawn | queued)
retention.retentionDays / retention.maxRuns	Per-pipeline run-history retention (requires global housekeeping config). See Run housekeeping.
delta.mode / delta.field / delta.dateField / delta.timeField	Watermark strategy for entity-shape reads
source.batchSize / source.delay / source.maxRetries / source.retryDelay	Paging and retry tuning
flags	Entity-cache truncate flags
description	UI label

Anything else is rejected (source.entity, target.*, viewMapping, source.query, refresh.slice, custom adapters, …). Structural changes require a code change and redeploy.

Live schedule changes

Engine	Live stop/restart without process restart?
spawn (interval)	Always — clearInterval + restart
queued (cron / CAP Event Queues)	Only when the runtime supports named-task unschedule (CDS 10+). On CDS 9 anonymous queues, change requires restart (or switch to spawn).

configView().meta.scheduleLiveChangeSupported surfaces this per pipeline so the console can warn before a doomed call.

API surface

On DataPipelineService (programmatic) and bound on /pipeline/Pipelines(...):

• setOverrides(patch) / clearOverrides(keys?) — merge or clear the JSON delta
• setEnabled(boolean) — convenience for the pause toggle
• setSchedule({ every, cron, engine }) / clearSchedule() — schedule overrides (persisted)
• configView() — { base, overrides, effective, fields[], meta } for the coded-vs-override-vs-effective diff

See Management Service for OData payloads and Pipeline Console for the UI.