Schedules & SLA

Schedule Configuration

The schedule block controls when and how the pipeline evaluates readiness. Schedules use cron expressions and support evaluation windows with configurable polling intervals.

schedule:
  cron: "0 8 * * *"
  timezone: UTC
  trigger:
    key: upstream-complete
    check: equals
    field: status
    value: ready
  evaluation:
    window: 1h
    interval: 5m

Schedule Fields

Trigger Condition

The trigger inside schedule defines the condition that activates the pipeline evaluation loop. This is separate from the validation rules that determine readiness.

schedule:
  trigger:
    key: upstream-complete
    check: equals
    field: status
    value: ready

When the trigger condition is met (e.g., an upstream pipeline signals completion), the evaluation loop begins. The loop then evaluates the full set of validation rules at the configured interval.

The trigger condition uses the same check syntax as validation rules: exists, equals, gt, gte, lt, lte, age_lt, age_gt.

Evaluation Window

The evaluation block controls how long and how often the pipeline checks its validation rules after the schedule activates.

schedule:
  evaluation:
    window: 1h
    interval: 5m

If the validation rules are not satisfied within the evaluation window, the Step Functions execution transitions to a timeout state and emits an EventBridge event.

Event-Driven Schedules

Pipelines can omit the cron field entirely and rely solely on sensor writes to trigger evaluation. When a sensor record is written to DynamoDB, the stream-router Lambda starts the Step Functions execution.

schedule:
  trigger:
    key: orders-landed
    check: exists
  evaluation:
    window: 30m
    interval: 2m

This pattern is useful for event-driven pipelines that should evaluate immediately when upstream data arrives.

Execution Granularity

Pipelines run at daily granularity by default. For sub-daily (hourly) execution, sensor data must include both date and hour fields. The framework builds a composite execution date (e.g., 2026-03-03T10) that scopes the entire execution — trigger lock, joblog entries, SLA, and retries — to that specific hour.

Glue triggers automatically receive --par_day (YYYYMMDD) and --par_hour (HH) arguments derived from the composite date.

The granularity is determined by the sensor data, not the pipeline config. A single pipeline can process both daily and hourly executions depending on what its sensors provide.

Cron-Based Schedules

For time-based schedules, specify a cron expression. The pipeline activates at the cron time, then evaluates until the window expires or validation passes.

schedule:
  cron: "0 8 * * *"
  timezone: America/New_York
  evaluation:
    window: 2h
    interval: 10m

Common cron patterns:

SLA Configuration

SLAs enforce time-based constraints. Breaching an SLA emits an EventBridge event but does not cancel the pipeline.

sla:
  deadline: "10:00"
  expectedDuration: 30m

SLA Fields

Deadline Evaluation

When the Step Functions execution reaches the SLA checkpoint, the sla-monitor Lambda creates two one-time EventBridge Scheduler entries:

• Warning: fires at deadline - expectedDuration
• Breach: fires at deadline

Each Scheduler entry invokes the sla-monitor Lambda at the exact timestamp to publish the corresponding event. Entries auto-delete after firing.

When the pipeline job completes (or all retries are exhausted), the state machine cancels any unfired Scheduler entries. If the job completed before the warning time, a SLA_MET event is published.

The deadline is interpreted in the schedule’s configured timezone (or UTC if not set).

EventBridge SLA Events

SLA events are published to the custom EventBridge bus. You can create rules to route these events to SNS, Lambda, or any other EventBridge target.

Example event:

{
  "source": "interlock",
  "detail-type": "SLA_BREACH",
  "detail": {
    "pipelineId": "gold-revenue",
    "scheduleId": "daily",
    "date": "2026-03-01",
    "message": "Pipeline gold-revenue: SLA_BREACH"
  }
}

Calendar Exclusions

Pipelines can be excluded from running on specific days of the week or calendar dates. When excluded, the pipeline is dormant – no evaluation, no SLA monitoring.

Inline Exclusions

exclusions:
  days:
    - saturday
    - sunday
  dates:
    - "2026-12-25"
    - "2026-01-01"

Named Calendar Reference

Define reusable calendars in YAML files and reference them from pipeline configs:

# calendars/holidays.yaml
name: holidays
days:
  - saturday
  - sunday
dates:
  - "2026-12-25"
  - "2026-01-01"
  - "2026-07-04"

Reference from a pipeline:

exclusions:
  calendar: holidays

Combining Calendar and Inline Exclusions

Inline days and dates are merged with the named calendar. A date or day matching any source triggers exclusion.

exclusions:
  calendar: holidays
  days:
    - friday
  dates:
    - "2026-03-15"

Calendar YAML files are loaded from the calendars_path directory configured in the Terraform module.