fix(cluster): read node readiness under check mode so the wait can judge it

roles/github_runner_cluster/tasks/node_readiness.yml

@@ -1,12 +1,13 @@
 ---
-# Reads, from the cluster itself, every node's Ready condition (github_runner_cluster_node_readiness, "<node>=<True|False|Unknown>" a line, the status empty for a node with no Ready condition yet) and the nodes memory-pressure regrading has stepped down (github_runner_cluster_nodes_stepped_down). Run when the answer is needed, never cached, since the rescue of tasks/wait_for_nodes.yml reads them after a wait that took a while.
+# Reads, from the cluster itself, every node's Ready condition (github_runner_cluster_node_readiness, "<node>=<True|False|Unknown>" a line, the status empty for a node with no Ready condition yet) and the nodes memory-pressure regrading has stepped down (github_runner_cluster_nodes_stepped_down). Run when the answer is needed, never cached, since the rescue of tasks/wait_for_nodes.yml reads them after a wait that took a while. Both reads change nothing, so they run under --check too: the wait that follows depends on their output.
 
 - name: Read which nodes are Ready
   ansible.builtin.command:
     argv: [kubectl, get, nodes, -o, 'jsonpath={range .items[*]}{.metadata.name}{"="}{.status.conditions[?(@.type=="Ready")].status}{"\n"}{end}']
   environment:
     KUBECONFIG: "{{ github_runner_cluster_kubeconfig_path }}"
   changed_when: false
+  check_mode: false
   register: github_runner_cluster_node_readiness
 
 - name: Read which nodes memory-pressure regrading has stepped down
@@ -16,4 +17,5 @@
   environment:
     KUBECONFIG: "{{ github_runner_cluster_kubeconfig_path }}"
   changed_when: false
+  check_mode: false
   register: github_runner_cluster_nodes_stepped_down

tests/unit/test_cluster_wait_nodes.py

@@ -77,9 +77,9 @@ def setUp(self) -> None:
         modules.mkdir(parents=True)
         (modules / "k8s.py").write_text(K8S_MODULE)
 
-    def wait(self) -> subprocess.CompletedProcess[str]:
+    def wait(self, check: bool = False) -> subprocess.CompletedProcess[str]:
         hosts, group_vars = self.hosts(("node1", "node2", "node3"), ("agent1",))
-        return self.run_tasks("wait_for_nodes", hosts, group_vars, limit="node1")
+        return self.run_tasks("wait_for_nodes", hosts, group_vars, check=check, limit="node1")
 
     def kubectl_calls(self) -> list[str]:
         return self.read_log("kubectl.log")
@@ -173,6 +173,15 @@ def test_a_cluster_holding_no_node_of_the_inventory_fails_rather_than_passing_an
         self.assertEqual(self.waits(), [])
 
 
+class CheckModeTest(WaitTestCase):
+    def test_a_check_run_reads_the_nodes_and_leaves_the_wait_itself_unrun(self) -> None:
+        self.nodes(node1="True", node2="True", node3="True", agent1="True")
+        result = self.wait(check=True)
+        self.assertEqual(result.returncode, 0, result.stdout)
+        self.assertEqual(self.waits(), [])
+        self.assertEqual(len([call for call in self.kubectl_calls() if call.startswith("get nodes")]), 2)
+
+
 class WiringTest(unittest.TestCase):
     def tasks(self, name: str) -> list[dict[str, Any]]:
         loaded: list[dict[str, Any]] = yaml.safe_load((ROLE / "tasks" / name).read_text())