Refactor logic in _is_node_in_replacement_valid() to account for `n…

…ode.instance` being `None`

src/slurm_plugin/clustermgtd.py

@@ -1207,17 +1207,17 @@ def _is_node_in_replacement_valid(self, node: SlurmNode, check_node_is_valid):
         If check_node_is_valid=True, check whether a node is in replacement,
         If check_node_is_valid=False, check whether a node is replacement timeout.
         """
-        if (
-            node.is_backing_instance_valid(
-                self._config.ec2_instance_missing_max_count,
-                self._nodes_without_backing_instance_count_map,
-                log_warn_if_unhealthy=True,
-            )
-            and node.name in self._static_nodes_in_replacement
-        ):
-            time_is_expired = time_is_up(
+        log.debug(f"Checking if node is in replacement {node}")
+        if node.is_backing_instance_valid(
+            self._config.ec2_instance_missing_max_count,
+            self._nodes_without_backing_instance_count_map,
+            log_warn_if_unhealthy=True,
+        ) and node.name in self._static_nodes_in_replacement:
+            # Set `time_is_expired` to `False` if `node.instance` is `None` since we don't have a launch time yet
+            time_is_expired = False if not node.instance else time_is_up(
                 node.instance.launch_time, self._current_time, grace_time=self._config.node_replacement_timeout
             )
+            log.debug(f"Node {node} is in replacement and timer expired? {time_is_expired}, instance? {node.instance}")
             return not time_is_expired if check_node_is_valid else time_is_expired
         return False
 

src/slurm_plugin/slurm_resources.py

@@ -474,7 +474,7 @@ def is_backing_instance_valid(
                 if log_warn_if_unhealthy:
                     logger.warning(
                         f"Incrementing missing EC2 instance count for node {self.name} to "
-                        f"{nodes_without_backing_instance_count_map[self.name]}."
+                        f"{nodes_without_backing_instance_count_map[self.name].count}."
                     )
         else:
             # Remove the slurm node from the map since the instance is healthy