homelab: runpod-lazy v5.2 manager-first boot, US-preference with global fallback

b850d799b7156380dacbff2229334b55b01e709f

permissionBRICK <40219477+permissionBRICK@users.noreply.github.com>

homelab/runpod-lazy-proxy.py+14 -9 Viewed

     return None, None, None
 def create_pod(values, datacenters=None):
     files = []
     for value in values or []:
         entry = catalog_entry(value)

         'containerDiskInGb': 80,
         'ports': ['8188/http', '8189/http'],
         'env': env,
 -        # The model-manager subshell tolerates old images without the script.
 +        # Manager starts FIRST so boot download progress is visible from
 -        'dockerStartCmd': ['bash', '-c', f'python3 /boot-models.py && (python3 /model-manager.py &) && cd /comfyui && exec python main.py {COMFY_ARGS}'],
 +        # outside; the subshell tolerates old images without the script.
 +        'dockerStartCmd': ['bash', '-c', f'(python3 /model-manager.py &) ; python3 /boot-models.py && cd /comfyui && exec python main.py {COMFY_ARGS}'],
+    }
     if DATACENTERSdatacenters:
         body['dataCenterIds'] = DATACENTERSdatacenters
     pod = api('POST', '/pods', body)
     log(f"pod created: {pod['id']} models={values_key(values) or 'legacy-all'} {pod.get('machine', {}).get('gpuTypeId')} ${pod.get('costPerHr')}/hr dc={pod.get('machine', {}).get('dataCenterId') or (','.join(datacenters) if datacenters else 'any')}")
     return pod['id'], pod.get('machine', {}).get('gpuTypeId')

     state['phase'] = 'orange'
     state['since'] = time.time()
     last_err = None
 -    for attempt in range(3):
 +    # Prefer the DATACENTERS list (e.g. US - fast HuggingFace peering for model
 +    # downloads); if it has no capacity, fall back to any datacenter.
 +    plans = [DATACENTERS, DATACENTERS, None] if DATACENTERS else [None, None, None]
 +    for attempt, datacenters in enumerate(plans):
         try:
             return create_pod(values, datacenters)
         except urllib.error.HTTPError as err:
             last_err = err.read().decode()[:200]
 -            log(f'create_pod attempt {attempt + 1} failed:', last_err)
 +            scope = 'preferred DCs' if datacenters else 'global'
 +            log(f'create_pod attempt {attempt + 1} ({scope}) failed:', last_err)
             time.sleep(5)
     state['phase'] = 'red'
     raise RuntimeError(f'could not create pod: {last_err}')