add missing availability zones (#11)

* add missing availability zones
Signed-off-by: vsoch <[email protected]>

fluxcloud/main/client.py

@@ -35,6 +35,7 @@ def run_timed(self, name, cmd, cleanup_func=None):
         """
         Run a timed command, and handle nonzero exit codes.
         """
+        logger.debug("\nRunning Timed Command:" + " ".join(cmd))
         res = utils.run_command(cmd)
 
         # An optional cleanup function (also can run if not successful)
@@ -54,6 +55,7 @@ def get_script(self, name, cloud=None):
         cloud = cloud or self.name
         script = os.path.join(here, "clouds", cloud, "scripts", name)
         if os.path.exists(script):
+            logger.debug(f"Found template script {script}")
             return script
 
     def get_shared_script(self, name):
@@ -81,6 +83,13 @@ def experiment_is_run(self, setup, experiment):
         # If all job output files exist, experiment is considered run
         for size in minicluster["size"]:
 
+            # We can't run if the minicluster > the experiment size
+            if size > experiment["size"]:
+                logger.warning(
+                    f"Cluster of size {experiment['size']} cannot handle a MiniCluster of size {size}, not considering."
+                )
+                continue
+
             # Jobname is used for output
             for jobname, _ in jobs.items():
 
@@ -151,6 +160,13 @@ def apply(self, setup, experiment):
         # NOTE if this changes here, also check self.experiment_is_run
         for size in minicluster["size"]:
 
+            # We can't run if the minicluster > the experiment size
+            if size > experiment["size"]:
+                logger.warning(
+                    f"Cluster of size {experiment['size']} cannot handle a MiniCluster of size {size}, skipping."
+                )
+                continue
+
             # Jobname is used for output
             for jobname, job in jobs.items():
 

fluxcloud/main/clouds/aws/client.py

@@ -111,7 +111,13 @@ def generate_config(self, setup, experiment):
         values["kubernetes_version"] = setup.settings.kubernetes["version"]
         values["size"] = setup.get_size(experiment)
         values["ssh_key"] = self.settings.aws.get("ssh_key")
-        values["availability_zones"] = self.settings.aws.get("availability_zones") or []
+        zones = self.settings.aws.get("availability_zones")
+
+        # If we don't have availability zones, provide a and b (min)
+        if not zones:
+            zones = ["%sa" % self.region, "%sb" % self.region]
+
+        values["availability_zones"] = zones
 
         # All extra custom variables
         values["variables"] = experiment.get("variables", {})

fluxcloud/main/clouds/aws/templates/cluster-config.yaml

@@ -9,7 +9,7 @@ metadata:
     "{{ tag[0] }}": "{{ tag[1] }}"
     {% endfor %}{% endif %}
 
-availabilityZones: [{% for zone in availability_zones %}"{{ zone }}"{% if loop.last %}{% else %}.{% endif %}{% endfor %}]
+availabilityZones: [{% for zone in availability_zones %}"{{ zone }}"{% if loop.last %}{% else %}, {% endif %}{% endfor %}]
 managedNodeGroups:
   - name: workers
     instanceType: {{ machine }}