Skip to main content

Manage Workloads

Overview​

Workloads are applications and jobs that run on your clusters. Bridge supports both custom workloads and pre-configured catalog workloads.

Create Custom Workload​

Custom Workload Definition​

Build your own workload from scratch:

  • Define containers
  • Configure resources
  • Set environment variables
  • Specify storage needs

Step 1: Create Workload​

Navigate to Workloads → Create Custom Workload:

Create Custom Workload

Step 2: Configure Container​

Define your container:

  • Image - Docker image to use
  • Registry - Container registry (if private)
  • Command - Entry point command
  • Arguments - Command arguments

Container Configuration

Step 3: Set Resources​

Allocate resources:

resources:
  requests:
    cpu: "2"
    memory: "4Gi"
  limits:
    cpu: "4"
    memory: "8Gi"

Specify:

  • CPU cores
  • Memory amount
  • GPU allocation
  • Storage requirements

Resource Configuration

Step 4: Environment & Storage​

Set environment variables:

ENV_VAR_1=value1
ENV_VAR_2=value2

Configure storage:

  • Mount points
  • Persistent volumes
  • Storage size

Environment Configuration

Step 5: Deploy Workload​

Review configuration and deploy:

Deploy Custom Workload

Workload will be:

  1. Scheduled on cluster
  2. Pulled from registry
  3. Started with specified configuration
  4. Monitored for health

Create Catalog Workload​

Pre-Built Templates​

Use pre-configured workload templates:

  • TensorFlow training
  • PyTorch development
  • Data processing pipelines
  • Web applications

Step 1: Browse Catalog​

Navigate to Workloads → Catalog:

Workload Catalog

Step 2: Select Template​

Choose a workload template:

Select Template

Review template details:

  • Description
  • Resource requirements
  • Configuration options
  • Input/output specifications

Step 3: Customize Workload​

Modify template parameters:

  • Resource allocation
  • Environment variables
  • Input data location
  • Output directory

Customize Workload

Step 4: Deploy Workload​

Deploy the customized workload:

Deploy Catalog Workload

Monitor Workloads​

View Workload Status​

Check running workloads:

Workload Status

Status indicators:

  • Running - Workload is executing
  • Pending - Waiting for resources
  • Completed - Job finished successfully
  • Failed - Workload encountered error

View Logs​

Access workload logs:

kubectl logs <workload-name>

Or through UI:

Workload Logs

View:

  • Standard output
  • Error messages
  • Application logs
  • System events

Monitor Metrics​

Track workload performance:

Workload Metrics

  • CPU usage
  • Memory usage
  • GPU usage
  • Network I/O
  • Disk I/O

Delete Workload​

Stop Workload​

Stop running workload:

  1. Select workload
  2. Click Stop
  3. Confirm stopping

Stop Workload

Delete Workload​

Remove workload permanently:

  1. Select workload
  2. Click Delete
  3. Confirm deletion

Delete Workload

Resources will be freed and returned to your quota.

Best Practices​

Workload Design​

  • Use container best practices
  • Include health checks
  • Handle signals gracefully
  • Log to standard output

Resource Management​

  • Request only needed resources
  • Set appropriate limits
  • Monitor actual usage
  • Adjust over time

Error Handling​

  • Implement retry logic
  • Handle timeouts
  • Clean up resources on exit
  • Log errors effectively

Common Workloads​

Data Processing​

  • ETL pipelines
  • Data transformation
  • Batch processing

Machine Learning​

  • Model training
  • Hyperparameter tuning
  • Feature engineering

Web Services​

  • APIs
  • Web applications
  • Microservices

Next Steps​

Last updated on