gpu-dev 0.5.23__tar.gz → 0.5.24__tar.gz

{gpu_dev-0.5.23 → gpu_dev-0.5.24}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: gpu-dev
-Version: 0.5.23
+Version: 0.5.24
 Summary: CLI tool for PyTorch GPU developer server reservations
 Author: PyTorch Team
 Requires-Python: >=3.10

{gpu_dev-0.5.23 → gpu_dev-0.5.24}/cli-tools/gpu-dev-cli/gpu_dev.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: gpu-dev
-Version: 0.5.23
+Version: 0.5.24
 Summary: CLI tool for PyTorch GPU developer server reservations
 Author: PyTorch Team
 Requires-Python: >=3.10

{gpu_dev-0.5.23 → gpu_dev-0.5.24}/cli-tools/gpu-dev-cli/gpu_dev_cli/auth.py

@@ -38,6 +38,19 @@ def _load_auth_cache(github_user: str) -> Optional[Dict[str, Any]]:
             return None
         if time.time() - float(entry.get("ts", 0)) > _AUTH_CACHE_TTL_SECONDS:
             return None
+        # Defense against stale cache on a persistent disk that pre-dates the IRSA fix:
+        # if AWS_ROLE_ARN points at a role the cached ARN doesn\'t reference, the cache
+        # is from a different identity (e.g. IMDS-fallback before fs_group=1081 landed)
+        # and should be ignored.
+        expected_role_arn = os.environ.get("AWS_ROLE_ARN", "")
+        cached_arn = (entry.get("result") or {}).get("arn", "")
+        if expected_role_arn:
+            try:
+                role_name = expected_role_arn.rsplit("/", 1)[-1]
+                if role_name and role_name not in cached_arn:
+                    return None
+            except Exception:
+                pass
         return entry.get("result")
     except Exception:
         return None

{gpu_dev-0.5.23 → gpu_dev-0.5.24}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "gpu-dev"
-version = "0.5.23"
+version = "0.5.24"
 description = "CLI tool for PyTorch GPU developer server reservations"
 authors = [{name = "PyTorch Team"}]
 readme = "cli-tools/gpu-dev-cli/README.md"

{gpu_dev-0.5.23 → gpu_dev-0.5.24}/terraform-gpu-devservers/lambda/reservation_processor/index.py

@@ -2888,6 +2888,7 @@ def allocate_gpu_resources(reservation_id: str, request: dict[str, Any], trace_d
             gpu_count=gpu_count,
             gpu_type=gpu_type,
             github_public_key=github_public_key,
+            github_user=github_user,
             reservation_id=reservation_id,
             jupyter_enabled=jupyter_enabled,
             persistent_volume_id=persistent_volume_id,
@@ -3434,7 +3435,8 @@ def create_kubernetes_resources(
     gpu_count: int,
     gpu_type: str,
     github_public_key: str,
-    reservation_id: str,
+    github_user: str = "",
+    reservation_id: str = None,
     jupyter_enabled: bool = False,
     persistent_volume_id: str = None,
     user_id: str = None,
@@ -3538,6 +3540,7 @@ def create_kubernetes_resources(
                         gpu_count,
                         gpu_type,
                         github_public_key,
+                        github_user=github_user,
                         jupyter_enabled=True,
                         persistent_volume_id=persistent_volume_id,
                         user_id=user_id,
@@ -3627,6 +3630,7 @@ def create_kubernetes_resources(
                         gpu_count,
                         gpu_type,
                         github_public_key,
+                        github_user=github_user,
                         jupyter_enabled=False,
                         persistent_volume_id=persistent_volume_id,
                         user_id=user_id,
@@ -3979,6 +3983,7 @@ def create_pod(
     gpu_count: int,
     gpu_type: str,
     github_public_key: str,
+    github_user: str = "",
     jupyter_enabled: bool = False,
     persistent_volume_id: str = None,
     user_id: str = None,

{gpu_dev-0.5.23 → gpu_dev-0.5.24}/terraform-gpu-devservers/lambda.tf

@@ -180,7 +180,7 @@ 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                     = "0.5.24"
+      LAMBDA_VERSION                     = "0.5.25"
       MIN_CLI_VERSION                    = "0.5.16"
       DISK_CONTENTS_BUCKET               = aws_s3_bucket.disk_contents.bucket
       OPERATIONS_TABLE                   = aws_dynamodb_table.operations.name