Reject negative retry durations (#3885)

Author: pragnyanramtha

src/dstack/_internal/server/services/runs/spec.py

@@ -4,6 +4,7 @@
     SERVICE_HTTPS_DEFAULT,
     ServiceConfiguration,
 )
+from dstack._internal.core.models.profiles import ProfileRetry
 from dstack._internal.core.models.repos.virtual import DEFAULT_VIRTUAL_REPO_ID, VirtualRunRepoData
 from dstack._internal.core.models.routers import RouterType
 from dstack._internal.core.models.runs import LEGACY_REPO_DIR, AnyRunConfiguration, RunSpec
@@ -74,6 +75,7 @@ def validate_run_spec_and_set_defaults(
     # the defaults depend on the server version, not the client version.
     if run_spec.run_name is not None:
         validate_dstack_resource_name(run_spec.run_name)
+    _validate_retry_duration(run_spec)
     for mount_point in run_spec.configuration.volumes:
         if not is_valid_docker_volume_target(mount_point.path):
             raise ServerClientError(f"Invalid volume mount path: {mount_point.path}")
@@ -132,6 +134,12 @@ def validate_run_spec_and_set_defaults(
         run_spec.configuration.working_dir = LEGACY_REPO_DIR
 
 
+def _validate_retry_duration(run_spec: RunSpec) -> None:
+    retry = run_spec.merged_profile.retry
+    if isinstance(retry, ProfileRetry) and retry.duration is not None and retry.duration < 0:
+        raise ServerClientError("retry.duration cannot be negative")
+
+
 def _check_dynamo_in_place_update_compatibility(
     current_run_spec: RunSpec, new_run_spec: RunSpec
 ) -> None:

src/tests/_internal/server/services/runs/test_spec.py

@@ -1,16 +1,19 @@
 import re
 import uuid
+from types import SimpleNamespace
 
 import pytest
 
 from dstack._internal.core.errors import ServerClientError
 from dstack._internal.core.models.configurations import ServiceConfiguration
 from dstack._internal.core.models.files import FileArchiveMapping
+from dstack._internal.core.models.profiles import Profile, ProfileRetry
 from dstack._internal.core.models.repos.local import LocalRunRepoData
 from dstack._internal.core.models.runs import RunSpec
 from dstack._internal.server.services.runs.spec import (
     _check_can_update_configuration,
     check_can_update_run_spec,
+    validate_run_spec_and_set_defaults,
 )
 from dstack._internal.server.testing.common import get_run_spec
 
@@ -77,6 +80,24 @@ def _run_spec_with_overrides(configuration: ServiceConfiguration, **overrides) -
     return RunSpec.parse_obj({**run_spec.dict(), **run_spec_overrides})
 
 
+class TestValidateRunSpecRetryDuration:
+    def test_model_accepts_negative_retry_duration_for_backward_compatibility(self):
+        retry = ProfileRetry(duration=-1)
+
+        assert retry.duration == -1
+
+    def test_rejects_negative_retry_duration_for_new_run_specs(self):
+        run_spec = get_run_spec(
+            repo_id="test-repo",
+            profile=Profile(name="default", retry=ProfileRetry(duration=-1)),
+        )
+
+        with pytest.raises(ServerClientError, match="retry.duration cannot be negative"):
+            validate_run_spec_and_set_defaults(
+                SimpleNamespace(ssh_public_key="ssh-rsa test"), run_spec
+            )
+
+
 class TestCheckCanUpdateConfigurationRouterType:
     def test_sglang_to_dynamo_router_type_change_is_rejected(self):
         current = _run_spec(_service_configuration(router_type="sglang"))