fix: show upgrade message when backend resources don't exist

cli-tools/gpu-dev-cli/gpu_dev_cli/auth.py

@@ -3,7 +3,7 @@
 import subprocess
 import re
 from typing import Dict, Any
-from .config import Config
+from .config import Config, CLI_UPGRADE_MESSAGE
 from rich.spinner import Spinner
 
 
@@ -34,6 +34,9 @@ def authenticate_user(config: Config) -> Dict[str, Any]:
         }
 
     except Exception as e:
+        # Let upgrade messages pass through without re-wrapping
+        if CLI_UPGRADE_MESSAGE in str(e):
+            raise
         raise RuntimeError(f"AWS authentication failed: {e}")
 
 

cli-tools/gpu-dev-cli/gpu_dev_cli/config.py

@@ -3,9 +3,20 @@
 import os
 import json
 import boto3
+from botocore.exceptions import ClientError
 from pathlib import Path
 from typing import Dict, Any, Optional
 
+# Upgrade message shown when backend infrastructure has changed
+CLI_UPGRADE_MESSAGE = """
+The GPU Dev service has been updated and requires a newer CLI version.
+
+Please upgrade:
+  pip install --upgrade git+https://github.com/pytorch/osdc.git@release
+
+For more info: https://github.com/pytorch/osdc
+""".strip()
+
 
 class Config:
     """Zero-config AWS-based configuration"""
@@ -101,6 +112,13 @@ def get_queue_url(self) -> str:
         try:
             response = self.sqs_client.get_queue_url(QueueName=self.queue_name)
             return response["QueueUrl"]
+        except ClientError as e:
+            error_code = e.response.get("Error", {}).get("Code", "")
+            if error_code == "AWS.SimpleQueueService.NonExistentQueue":
+                raise RuntimeError(CLI_UPGRADE_MESSAGE)
+            raise RuntimeError(
+                f"Cannot access SQS queue {self.queue_name}. Check AWS permissions: {e}"
+            )
         except Exception as e:
             raise RuntimeError(
                 f"Cannot access SQS queue {self.queue_name}. Check AWS permissions: {e}"

cli-tools/gpu-dev-cli/gpu_dev_cli/reservations.py

@@ -16,7 +16,7 @@
 from rich.live import Live
 from rich.spinner import Spinner
 
-from .config import Config
+from .config import Config, CLI_UPGRADE_MESSAGE
 from .name_generator import sanitize_name
 from . import __version__
 
@@ -971,6 +971,13 @@ def get_gpu_availability_by_type(self) -> Optional[Dict[str, Dict[str, Any]]]:
 
             return availability_info
 
+        except ClientError as e:
+            error_code = e.response.get("Error", {}).get("Code", "")
+            if error_code == "ResourceNotFoundException":
+                raise RuntimeError(CLI_UPGRADE_MESSAGE)
+            console.print(
+                f"[red]❌ Error getting GPU availability: {str(e)}[/red]")
+            return None
         except Exception as e:
             console.print(
                 f"[red]❌ Error getting GPU availability: {str(e)}[/red]")

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "gpu-dev-cli"
-version = "0.3.5"
+version = "0.3.6"
 description = "CLI tool for PyTorch GPU developer server reservations"
 authors = [{name = "PyTorch Team"}]
 readme = "cli-tools/gpu-dev-cli/README.md"

terraform-gpu-devservers/lambda.tf

@@ -179,8 +179,8 @@ resource "aws_lambda_function" "reservation_processor" {
       HOSTED_ZONE_ID                     = local.effective_domain_name != "" ? local.hosted_zone_id : ""
       SSH_DOMAIN_MAPPINGS_TABLE          = local.effective_domain_name != "" ? aws_dynamodb_table.ssh_domain_mappings.name : ""
       SSL_CERTIFICATE_ARN                = local.effective_domain_name != "" ? aws_acm_certificate.wildcard[0].arn : ""
-      LAMBDA_VERSION                     = "0.3.5"
-      MIN_CLI_VERSION                    = "0.3.5"
+      LAMBDA_VERSION                     = "0.3.6"
+      MIN_CLI_VERSION                    = "0.3.6"
       DISK_CONTENTS_BUCKET               = aws_s3_bucket.disk_contents.bucket
     }, local.alb_env_vars)
   }