fix(kuberay): strip trailing non-alphanumeric chars after truncating K8s label values

`normalize_k8s_label_values` stripped leading/trailing non-alphanumeric
characters before truncating to 63 chars. For long inputs (e.g. git
branch names used as `dagster/git-branch`-style labels), truncation
could then land on `_`, `-`, or `.` and the resulting label would fail
K8s's label-value regex `(([A-Za-z0-9][-A-Za-z0-9_.]*)?[A-Za-z0-9])?`,
causing RayCluster creation to fail with HTTP 422.

Swap the order: truncate first, then strip. Add parametrized regression
tests covering trailing `_`, `.`, `-`, and a collapsed-hyphen run
straddling the 63-char boundary.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: peterroelants

src/dagster_ray/kuberay/utils.py

@@ -58,11 +58,13 @@ def normalize_k8s_label_values(labels: dict[str, str]) -> dict[str, str]:
         # Collapse multiple consecutive hyphens into a single hyphen
         value = re.sub(r"-+", "-", value)
 
-        # Remove leading/trailing non-alphanumeric characters
-        value = _LEADING_TRAILING_NON_ALNUM_PATTERN.sub("", value)
-
-        # Truncate to 63 characters
-        normalized[key] = value[:63]
+        # Truncate to 63 characters first, then strip leading/trailing
+        # non-alphanumeric characters. The order matters: stripping before
+        # truncating leaves the last char exposed to whatever lands at
+        # position 63 — which can be `_`, `-`, or `.` for long inputs and
+        # would then violate K8s's label-value regex (must end alphanumeric).
+        value = value[:63]
+        normalized[key] = _LEADING_TRAILING_NON_ALNUM_PATTERN.sub("", value)
 
     # replace keys starting with `dagster.io/` with `dagster/`
     normalized_with_fixed_dagster_keys = {}

tests/kuberay/test_utils.py

@@ -1,5 +1,13 @@
+import re
+
+import pytest
+
 from dagster_ray.kuberay.utils import normalize_k8s_label_values
 
+# From the K8s label-value spec:
+# https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/#syntax-and-character-set
+_K8S_LABEL_VALUE = re.compile(r"^(([A-Za-z0-9][-A-Za-z0-9_.]*)?[A-Za-z0-9])?$")
+
 
 def test_normalize_k8s_label_values(snapshot):
     assert snapshot == normalize_k8s_label_values(
@@ -18,3 +26,31 @@ def test_normalize_k8s_label_values(snapshot):
 
 def test_normalize_k8s_label_values_important_labels():
     assert normalize_k8s_label_values({"dagster/run-id": "12345"}) == {"dagster/run-id": "12345"}
+
+
+@pytest.mark.parametrize(
+    "value",
+    [
+        # Git branch name > 63 chars whose 63rd char is `_`. Regression for
+        # KubeRay RayCluster creation failing with
+        # `metadata.labels: Invalid value … must start and end with an
+        # alphanumeric character` when a PR's long branch name was being
+        # stuck into a `dagster/git-branch`-style label.
+        "04-22-datastack_reshape_metaxy_map-entry_columns_on_inputmodel_"
+        "to_silence_spurious_pydantic_serializer_warnings_in_ray_workers",
+        # Truncation landing on `.`
+        "a" * 63 + "." + "b" * 10,
+        # Truncation landing on `-`
+        "a" * 63 + "-" + "b" * 10,
+        # Collapsed-hyphen run exactly at position 63.
+        "a" * 62 + "--" + "b" * 10,
+    ],
+)
+def test_normalize_k8s_label_values_truncated_tail_is_alphanumeric(value: str) -> None:
+    """Truncation to 63 chars must never leave a trailing non-alphanumeric
+    character; doing so produces a label value that K8s rejects with HTTP 422."""
+    out = normalize_k8s_label_values({"k": value})["k"]
+
+    assert len(out) <= 63
+    assert out == "" or out[-1].isalnum(), f"label ends non-alphanumeric: {out!r}"
+    assert _K8S_LABEL_VALUE.match(out), f"value violates K8s label regex: {out!r}"