Worker Labels

Worker labels allow you to tag workers with specific capabilities and route tasks to appropriate workers based on their requirements.

TIP

Worker labels are a key component of Dagu's Distributed Execution feature. Make sure to understand the distributed execution architecture before implementing worker labels.

Configuring Worker Labels

Command Line

Specify labels when starting a worker using the --worker-labels flag:

# GPU-enabled worker
dagu worker --worker.labels gpu=true,memory=64G,region=us-east-1

# CPU-optimized worker  
dagu worker --worker.labels cpu-arch=amd64,cpu-cores=16,instance-type=m5.large

# Region-specific worker
dagu worker --worker.labels region=eu-west-1,compliance=gdpr

Configuration File

Set labels in the configuration file:

# config.yaml
worker:
  labels:
    gpu: "true"
    memory: "64G"
    region: "us-east-1"

Environment Variable

Set labels via environment variable:

export DAGU_WORKER_LABELS="gpu=true,memory=64G"
dagu worker

Using Worker Selectors in DAGs

Specify workerSelector on any step to route it to workers with matching labels:

steps:
  # This task will only run on workers with gpu=true label
  - name: train-model
    run: train

---
name: train
workerSelector:
  gpu: "true"
steps:
  - name: train
    command: python train.py

Label Matching Rules

1. All labels must match: A worker must have ALL labels specified in the workerSelector to be eligible
2. Empty selector: Tasks without workerSelector can run on any worker
3. Exact match: Label values must match exactly (case-sensitive)

Example Use Cases

GPU/CPU Task Routing

# GPU worker
dagu worker --worker.labels gpu=true

# CPU worker  
dagu worker --worker.labels cpu=true

# DAG
steps:
  - name: gpu-task
    run: gpu-task

  - name: cpu-task
    run: cpu-task

---
# Run on a worker with gpu
name: gpu-task
workerSelector:
  gpu-task: "true"
steps:
  - name: gpu task
    command: python gpu-task.py

---
# Run on a worker with faster cpu
name: cpu-task
workerSelector:
  cpu-task: "true"
steps:
  - name: cpu task
    command: python cpu-task.py

See Also

• Distributed Execution - Complete guide to distributed execution
• YAML Reference - workerSelector configuration
• CLI Reference - Worker command options
• Configuration Reference - Worker configuration