Managing Pods

Learn how to monitor, control, and manage your running pods.

Viewing Pods

Navigate to Compute > Pods to see all pods in the current project.

Pod List View

The list shows:

• Name: Pod identifier
• Status: Current state (Running, Stopped, etc.)
• Image: Container image being used
• Resources: GPU, CPU, memory allocation
• Age: Time since creation
• Cost: Current hourly rate

Filtering and Sorting

• Filter by status (Running, Stopped, All)
• Filter by project
• Search by pod name

Pod Actions

Start a Pod

For stopped pods:

1. Find the pod in the list
2. Click the Start button (play icon)
3. Wait for status to change to Running

Billing resumes when the pod starts.

Stop a Pod

To pause a running pod:

1. Click the Stop button (pause icon)
2. Confirm the action
3. Pod enters Stopped state

Benefits of stopping:

• Billing pauses immediately
• Pod configuration is preserved
• Can be restarted later

Note: Ephemeral storage may be lost when stopping. Use NFS volumes for persistent data.

Restart a Pod

To restart without stopping:

1. Click the Restart button
2. Container will stop and start again

Useful for applying configuration changes or recovering from issues.

Delete a Pod

To permanently remove a pod:

1. Click the Delete button (trash icon)
2. Confirm the deletion
3. Pod and its ephemeral storage are removed

Warning: Deletion is permanent. Ensure you’ve saved any important data.

Monitoring

Pod Details

Click on a pod to view detailed information:

Overview

• Full configuration details
• Creation time
• SSH connection string
• Jupyter URL (if enabled)
• Port mappings and access URLs

Stats (Real-Time) Live metrics streamed via WebSocket:

• CPU usage percentage
• Memory usage and limits
• GPU utilization per GPU
• GPU memory usage
• Network I/O (bytes in/out)
• Disk usage

Stats update automatically without refreshing the page.

Logs (Real-Time Streaming) View container output with live streaming:

• Real-time log streaming via WebSocket
• Logs appear as they’re generated
• Search and filter within logs
• Download logs for offline analysis
• Automatic reconnection on connection drop

Terminal Access an interactive terminal directly in your browser:

• Click Terminal button on pod details
• Opens in a new window for focused work
• Full terminal emulation with copy/paste support
• No SSH client or keys required
• Works through firewalls

Exposed Ports Manage exposed ports on running pods:

• View all currently exposed ports with their HTTPS URLs
• Add Port: Expose a new container port dynamically
• Remove Port: Stop exposing a port
• Each port gets an auto-generated HTTPS endpoint: https://<subdomain>-<port>.cloud.podstack.ai
• Port 22 (SSH) uses the format: ssh-<subdomain>.cloud.podstack.ai

Billing Information

Each pod detail page shows:

• Current hourly rate
• Total cost since creation
• Cost breakdown by resource type

Scaling Pods

For pods configured with replicas:

1. Go to pod details
2. Click Scale
3. Enter new replica count (increase or decrease)
4. Confirm the scaling operation

Scaling Behavior:

• Scaling up creates additional identical replicas
• Scaling down terminates excess replicas gracefully
• Each replica runs independently and is billed separately
• Billing adjusts immediately based on replica count

Use Cases:

• Scale up for parallel processing workloads
• Scale down during off-peak hours to save costs
• Dynamically adjust based on workload demands

Auto-Stop on Insufficient Balance

When your wallet balance becomes insufficient:

1. Warning notification is sent
2. Pods may be automatically stopped after a grace period
3. Data on ephemeral storage may be lost
4. NFS-mounted volumes preserve data

To Prevent Auto-Stop:

• Maintain adequate wallet balance
• Set up auto-debit for automatic top-ups
• Monitor low balance warnings
• Stop unused pods manually to conserve balance

Bulk Operations

To manage multiple pods:

1. Select pods using checkboxes
2. Use bulk action buttons (Stop All, Delete Selected)

Pod Health

Podstack monitors pod health automatically:

Check logs for details when health is degraded.

Best Practices

Cost Management

• Stop pods when not actively using them
• Delete pods when work is complete
• Use templates for reproducible deployments

Data Persistence

• Mount NFS volumes for important data
• Don’t rely on ephemeral storage for critical files
• Back up data before deleting pods

Resource Optimization

• Right-size GPU, CPU, and memory
• Monitor actual usage vs. allocation
• Reduce resources if over-provisioned

Next Steps

Learn about Connecting to Pods for SSH, terminal, and notebook access.