REST API: Pods show RUNNING but never start (uptime 0, no ports)

[zdavatz]

RunPod Bug Report: Pods show "RUNNING" but never actually start

Summary

Pods created via REST API show desiredStatus: "RUNNING" but never actually boot. uptimeInSeconds stays at 0, no ports are assigned, no SSH access is possible. The pod consumes credits but does nothing.

Tested Configurations

Test 1: RTX 5090, Community, pytorch image (2026-03-18 15:40 UTC)

Request:

curl -X POST "https://rest.runpod.io/v1/pods" \
  -H "Authorization: Bearer $RUNPOD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "davaz-test-5090",
    "gpuTypeIds": ["NVIDIA GeForce RTX 5090"],
    "gpuCount": 1,
    "imageName": "pytorch/pytorch:2.7.0-cuda12.8-cudnn9-runtime",
    "containerDiskInGb": 50,
    "volumeInGb": 200,
    "ports": ["22/tcp", "8080/http"],
    "cloudType": "COMMUNITY",
    "dockerStartCmd": ["bash", "-c", "sleep infinity"]
  }'

Create Response (success):

{
  "id": "xbwp04efggmm9i",
  "desiredStatus": "RUNNING",
  "costPerHr": 0.69,
  "gpuCount": 1,
  "imageName": "pytorch/pytorch:2.7.0-cuda12.8-cudnn9-runtime",
  "machine": {
    "gpuTypeId": "NVIDIA GeForce RTX 5090",
    "location": "CA",
    "diskThroughputMBps": 7300,
    "maxDownloadSpeedMbps": 7988
  },
  "machineId": "8ddg10q4fe7n",
  "memoryInGb": 125,
  "vcpuCount": 64,
  "volumeInGb": 200
}

Status after 30 seconds:

{
  "desiredStatus": "RUNNING",
  "runtime": {
    "uptimeInSeconds": 0,
    "ports": []
  },
  "publicIp": ""
}

Status after 90 seconds: Identical — uptime still 0, no ports.

Pod was manually destroyed to stop billing.

Test 2: RTX 5090, ubuntu:22.04 image (2026-03-18 09:50 UTC)

Request:

curl -X POST "https://rest.runpod.io/v1/pods" \
  -H "Authorization: Bearer $RUNPOD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "davaz-5090-bare",
    "gpuTypeIds": ["NVIDIA GeForce RTX 5090"],
    "gpuCount": 1,
    "imageName": "ubuntu:22.04",
    "containerDiskInGb": 20,
    "ports": ["22/tcp"]
  }'

Result: Same issue — desiredStatus: "RUNNING", uptimeInSeconds: 0, no ports assigned. Pod ran for hours doing nothing, consuming $0.89/hr.

Test 3: RTX PRO 6000, multiple datacenters (2026-03-18, earlier tests)

Same issue with:

• GPU: NVIDIA RTX PRO 6000 Blackwell Server Edition
• Datacenters: EU-RO-1, US-KS-2
• Images: pytorch, nvidia/cuda, ubuntu:22.04
• With and without network volumes

All pods stuck at uptime 0.

Network Volume Issues

Network volumes create successfully but pods using them also fail to start:

POST /v1/pods with "networkVolumeId": "fh1hx9ujcu"
→ {"error": "create pod: There are no instances currently available"}

POST /v1/pods with "networkVolumeId": "3vdttbgap9" (US-KS-2)
→ {"error": "create pod: could not find any pods with required specifications"}

When pods DO get assigned (without network volume, community cloud), they show RUNNING but never boot.

Environment

• API Key: Works correctly (can list pods, create/delete volumes, create pods)
• Account: Has sufficient credit ($5+ balance confirmed)
• SSH Key: Added to account settings at https://console.runpod.io/user/settings
• Date: 2026-03-18
• All tests via REST API v1 (https://rest.runpod.io/v1/)

Expected Behavior

Pod should transition from RUNNING to actually running with:

• uptimeInSeconds > 0
• SSH port assigned in runtime.ports
• Container accessible via SSH

Actual Behavior

• Pod shows desiredStatus: "RUNNING" immediately after creation
• runtime.uptimeInSeconds stays at 0 indefinitely
• runtime.ports stays empty []
• publicIp stays empty
• No SSH access possible
• Pod consumes credits despite not running
• Destroying and recreating does not help
• Different GPUs, images, datacenters, and configurations all fail the same way

Impact

• Cannot use RunPod for any GPU workload
• Credits consumed by non-functional pods (~$0.89/hr per pod)
• Multiple pods created and destroyed over the day, all with same result
• Competing platforms (vast.ai, TensorDock) work correctly with identical workloads