Delay (Scheduled pause)

Description

The Delay node halts a workflow for a configured amount of time (seconds, minutes, hours, or days) and resumes execution automatically when the deadline is reached, continuing from the next configured node. Unlike a blocking sleep, the Delay frees the server process during the wait: the state is persisted to the database and the workflow resumes in a new process when the scheduled time arrives.

The execution trace is continuous: nodes run before and after the delay appear in the same instance history.

Configuration

Field	Type	Required	Description
delayAmount	Number	Yes	How long to wait
delayUnit	Select	Yes	Unit: seconds, minutes, hours, days
nextModule	Module	Yes	Node that will execute when the delay expires

How it works

1. The node calculates the exact resume date/time (now + delayAmount × delayUnit).
2. Persists the continuation state in the workflow_delays table with status = waiting.
3. Schedules a node-schedule job for that date.
4. Returns _stopflowprocess: true to the engine, which pauses the flow and releases the execution slot.
5. When the job fires, continueWorkflowAfterDelay calls the engine with the same instance workflow_id and the next node as the entry point.
6. The engine detects the instance already exists, loads the steps recorded before the delay from the DB, and pre-populates the trace array — ensuring full session continuity.

Recovery after server restarts

If the server restarts while delays are pending, DelayMonitorService restores them automatically on startup: it reschedules jobs whose execution date has not yet passed and immediately executes any that expired during downtime.

workflow_delays table

CREATE TABLE workflow_delays (
    id INT AUTO_INCREMENT PRIMARY KEY,
    workflow_id VARCHAR(255) NOT NULL,
    client_id INT NOT NULL,
    next_module VARCHAR(255) NOT NULL,
    continuation_data TEXT NOT NULL,
    scheduled_at DATETIME NOT NULL,
    delay_node_id VARCHAR(255) NOT NULL,
    status ENUM('waiting', 'executed', 'error', 'cancelled') DEFAULT 'waiting',
    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
    executed_at TIMESTAMP NULL,
    error_message TEXT NULL,
    UNIQUE KEY unique_workflow_delay (workflow_id, delay_node_id)
);

Output while waiting

When the Delay node activates, it immediately returns this object to the engine (the flow stops here):

{
  "nextModule": null,
  "_stopflowprocess": true,
  "_meta_": {
    "delayScheduled": true,
    "executeAt": "2025-06-10T14:30:00.000Z",
    "jobName": "delay_wf_abc123_nodeId"
  },
  "data": {
    "...": "flow data",
    "delayInfo": {
      "scheduledAt": "2025-06-10T14:30:00.000Z",
      "delayAmount": 30,
      "delayUnit": "minutes",
      "status": "scheduled"
    }
  }
}

Possible errors

Code	Description
MISSING_TABLE	The workflow_delays table does not exist in the database
INVALID_TABLE_STRUCTURE	The table exists but is missing required columns
INVALID_DELAY_AMOUNT	delayAmount is not a positive number
INVALID_DELAY_UNIT	delayUnit is not one of the allowed values
MISSING_NEXT_MODULE	No next node was configured
INVALID_EXECUTION_TIME	The calculated date is already in the past
SCHEDULE_ERROR	Internal error scheduling the job

Usage examples

Retry an HTTP call after 5 minutes

{
  "delayAmount": 5,
  "delayUnit": "minutes"
}

Wait 1 day before sending a follow-up email

{
  "delayAmount": 1,
  "delayUnit": "days"
}

Separate two processing batches by 30 seconds

{
  "delayAmount": 30,
  "delayUnit": "seconds"
}

Notes

• The workflow does not hold an execution slot during the pause; other workflows run normally.
• If the same workflow_id + delay_node_id pair already has a registered delay, INSERT ... ON DUPLICATE KEY UPDATE replaces the previous record (the prior job is cancelled and a new one is created).
• The data flowing through the pipeline is serialized as JSON in continuation_data and restored on resume.
• The full trace (nodes before and after the delay) is visible in the instance history.

Related nodes

• SubWorkflow — executes a child flow synchronously while maintaining session linkage
• Iterator — splits an array into batches and processes them sequentially or in parallel