Operation

Operation

This guide covers day-to-day usage of Blazor Data Orchestrator — managing jobs, monitoring execution, and configuring the platform.

---

Job Lifecycle

---

Dashboard Overview

The Dashboard page displays all jobs with key status indicators:

Column	Description
Job Name	Name of the job
Enabled	Whether the job is active and eligible for scheduling
Queued	Whether the job is currently queued for execution
In Process	Whether an agent is currently executing the job
Error	Whether the last execution encountered an error

From the dashboard, you can:

• Click a job row to open the Job Details dialog
• Click Create Job to define a new job
• Use the Run Job Now quick action to trigger immediate execution

---

Creating a Job

1. Click Create Job on the home page.
2. Fill in the required fields:
   • Job Name — A descriptive name for the job
   • Job Groups (optional) — The organization this job belongs to
3. Click Save to create the job.

The new job is created in a disabled state. You need to upload or write code, configure a schedule, and enable the job before it will execute.

---

Job Details Dialog

The Job Details dialog is the central management interface for each job. It is organized into tabs:

Details Tab

• Base Job Name — View or edit the job name
• Enabled toggle — Enable or disable the job
• Environment — Set the execution environment
• Queue — Assign the job to a specific queue (e.g., default)
• Run Job Now button — Queue the job for immediate execution

Schedules Tab

Configure one or more schedules for the job:

• Schedule Name — A label for this schedule
• Enabled — Toggle the schedule on/off independently of the job
• Days — Select which days of the week the job should run (Mon–Sun)
• Start Time / Stop Time — Time window in military format (e.g., 800 to 1700)
• Run Every N Hours — Interval between runs within the time window

Parameters Tab

Attach key-value data to the job:

• Field Description — A label describing the parameter
• Value — Supports string, integer, or date values

Parameters are accessible to job code through the execution context.

Logs Tab

View execution history and log entries:

• Select a Job Instance from the dropdown to filter logs
• Each log entry shows: Action, Details, Level, and Timestamp
• Timestamps are adjusted by the configured timezone offset

Webhook Tab

Enable a webhook endpoint for the job:

• Toggle the Webhook Enabled switch
• A unique GUID is generated and displayed as the webhook URL
• External systems can trigger the job via GET or POST to /webhook/{GUID}

Code Tab

Write or upload job code. See the Online guide for detailed editor usage.

---

Running a Job

There are three ways to trigger job execution:

Method	Description
Scheduled	The Scheduler service evaluates enabled schedules at a configurable interval (default: 60 seconds). When conditions are met, it creates a JobInstance and queues a message.
Manual	Click Run Job Now on the Details tab or Code tab. This immediately creates a JobInstance and sends a queue message.
Webhook	An external system sends an HTTP request to /webhook/{GUID}. Query parameters are forwarded to the job.

---

Monitoring Execution

Viewing Logs

1. Navigate to the Logs tab.
2. Log entries are displayed in a table with columns for Action, Details, Level, and Timestamp.

Understanding Job Status

Status	Meaning
Queued	A message has been sent to the Azure Queue but no agent has picked it up yet
In Process	An agent has received the message and is executing the job
Completed	Execution finished successfully — the instance is marked complete
Error	Execution failed or the instance was marked as stuck by the scheduler

---

Administration

Navigate to the Administration page via the navigation link on the home page.

Job Groups Tab

• Create and manage organizational groups for categorizing jobs.
• Groups have a name and an active/inactive status.
• Assign jobs to groups for easier filtering and management.

Job Queues Tab

• Create and manage queue configurations.
• Each queue has a name (e.g., default, jobs-large-container, onprem-queue).
• Assign queues to jobs to route their execution to specific agent pools.
• Deploy agents configured with matching QueueName settings to process queue-specific workloads.

Timezone Settings Tab

• Timezone Offset — Configure the timezone offset (in hours) for log display. This offset is applied to all timestamps shown in the UI.

AI Settings Tab

Status	Meaning
AI Service Type	Choose the AI provider
API Key	Enter the API key supplied by the AI provider
Model	Choose the model to use

---

Agent Behavior

The Agent is a background worker service that processes job execution messages. It implements several production reliability patterns — queue-based scaling, NuGet packaging for job distribution, and heartbeat-based reliability for long-running tasks. Here is what happens when a job is executed:

1. Poll queue — The agent polls its configured queue every 5 seconds for new messages.
2. Receive message — When a message is received, the visibility timeout is set to 5 minutes.
3. Download package — The agent downloads the job's .nupkg file from Azure Blob Storage.
4. Extract & validate — The package is extracted and validated (checking for main.cs/main.py, configuration.json, etc.).
5. Resolve dependencies — For C# jobs, NuGet dependencies declared in the .nuspec are resolved using dotnet restore.
6. Load configuration — The appropriate appsettings file is loaded based on the job's environment setting. Connection strings from the agent's configuration are merged in.
7. Execute code — C# code is compiled and executed via Roslyn/CS-Script. Python code is executed via subprocess.
8. Log results — Execution logs are written to Azure Table Storage.
9. Update status — The JobInstance record is updated with completion or error status.
10. Delete message — On success, the queue message is deleted. On failure, it becomes visible again after the timeout.

Visibility Timeout Renewal (Heartbeat)

For jobs that run longer than 5 minutes, a background task renews the message visibility every 3 minutes. This heartbeat pattern is critical for production reliability:

• It prevents the message from becoming visible to other agents while the job is still running.
• If the agent crashes, the message becomes visible after the timeout, allowing another agent to pick it up — ensuring no job is silently lost.
• Combined with the Scheduler's stuck instance detection (which marks unresponsive instances as errors after a configurable timeout), this provides end-to-end reliability for long-running workloads.

Horizontal Scaling

Agents scale horizontally across multiple queues. You can:

• Scale replicas — Deploy multiple replicas of the same agent for parallel processing on a single queue.
• Create dedicated pools — Deploy separate agents with different QueueName values to create specialized processing pools (e.g., default for lightweight jobs, jobs-large-container for resource-intensive workloads).
• Mix environments — Run cloud-based agents alongside on-premises agents monitoring different queues.

---

Back to Home