Merge pull request #22 from hjacobs/ignore-master-asg

#21 ignore master nodes by default

Author: hjacobs

README.rst

@@ -49,7 +49,7 @@ The autoscaler consists of a simple main loop which calls the ``autoscale`` func
 The main loop keeps no state (like history), all input for the ``autoscale`` function comes from either static configuration or the Kubernetes API server.
 The ``autoscale`` function performs the following task:
 
-* retrieve the list of all nodes from the Kubernetes API and group them by Auto Scaling Group (ASG) and Availability Zone (AZ)
+* retrieve the list of all (worker) nodes from the Kubernetes API and group them by Auto Scaling Group (ASG) and Availability Zone (AZ)
 * retrieve the list of all pods from the Kubernetes API
 * calculate the current resource "usage" for every ASG and AZ by summing up all pod resource requests (CPU, memory and number of pods)
 * calculates the currently required number of nodes per AWS Auto Scaling Group:

kube_aws_autoscaler/main.py

@@ -80,7 +80,7 @@ def is_node_ready(node):
     return False
 
 
-def get_nodes(api) -> dict:
+def get_nodes(api, include_master_nodes: bool=False) -> dict:
     nodes = {}
     for node in pykube.Node.objects(api):
         region = node.labels['failure-domain.beta.kubernetes.io/region']
@@ -98,7 +98,8 @@ def get_nodes(api) -> dict:
                'ready': is_node_ready(node),
                'unschedulable': node.obj['spec'].get('unschedulable', False),
                'master': node.labels.get('master', 'false') == 'true'}
-        nodes[node.name] = obj
+        if include_master_nodes or not obj['master']:
+            nodes[node.name] = obj
     return nodes
 
 
@@ -317,10 +318,10 @@ def get_ready_nodes_by_asg(nodes_by_asg_zone):
     return ready_nodes_by_asg
 
 
-def autoscale(buffer_percentage: dict, buffer_fixed: dict, buffer_spare_nodes: int=0, dry_run: bool=False):
+def autoscale(buffer_percentage: dict, buffer_fixed: dict, buffer_spare_nodes: int=0, include_master_nodes: bool=False, dry_run: bool=False):
     api = get_kube_api()
 
-    all_nodes = get_nodes(api)
+    all_nodes = get_nodes(api, include_master_nodes)
     region = list(all_nodes.values())[0]['region']
 
     autoscaling = boto3.client('autoscaling', region)
@@ -344,6 +345,8 @@ def main():
     parser.add_argument('--debug', '-d', help='Debug mode: print more information', action='store_true')
     parser.add_argument('--once', help='Run loop only once and exit', action='store_true')
     parser.add_argument('--interval', type=int, help='Loop interval (default: 60s)', default=60)
+    parser.add_argument('--include-master-nodes', help='Do not ignore auto scaling group with master nodes',
+                        action='store_true')
     parser.add_argument('--buffer-spare-nodes', type=int,
                         help='Number of extra "spare" nodes to provision per ASG/AZ (default: 1)', default=1)
     for resource in RESOURCES:
@@ -368,7 +371,7 @@ def main():
     while True:
         try:
             autoscale(buffer_percentage, buffer_fixed, buffer_spare_nodes=args.buffer_spare_nodes,
-                      dry_run=args.dry_run)
+                      include_master_nodes=args.include_master_nodes, dry_run=args.dry_run)
         except:
             logger.exception('Failed to autoscale')
         if args.once:

tests/test_autoscaler.py

@@ -251,8 +251,21 @@ def test_get_nodes(monkeypatch):
         'spec': {'externalID': 'i-123'}
     }
 
+    master = MagicMock()
+    master.name = 'master'
+    master.labels = {
+        'failure-domain.beta.kubernetes.io/region': 'eu-north-1',
+        'failure-domain.beta.kubernetes.io/zone': 'eu-north-1a',
+        'beta.kubernetes.io/instance-type': 'a1.small',
+        'master': 'true'
+    }
+    master.obj = {
+        'status': {'allocatable': {'cpu': '2', 'memory': '16Gi', 'pods': '10'}},
+        'spec': {'externalID': 'i-456'}
+    }
+
     objects = MagicMock()
-    objects.return_value = [node]
+    objects.return_value = [node, master]
     monkeypatch.setattr('pykube.Node.objects', objects)
     api = MagicMock()
     assert get_nodes(api) == {'n1': {
@@ -263,6 +276,21 @@ def test_get_nodes(monkeypatch):
         'unschedulable': False,
         'master': False}}
 
+    assert get_nodes(api, include_master_nodes=True) == {'n1': {
+        'name': 'n1',
+        'region': 'eu-north-1', 'zone': 'eu-north-1a', 'instance_id': 'i-123', 'instance_type': 'x1.mega',
+        'allocatable': {'cpu': 2, 'memory': 16*1024*1024*1024, 'pods': 10},
+        'ready': False,
+        'unschedulable': False,
+        'master': False}, 'master': {
+            'name': 'master',
+            'region': 'eu-north-1', 'zone': 'eu-north-1a', 'instance_id': 'i-456', 'instance_type': 'a1.small',
+            'allocatable': {'cpu': 2, 'memory': 16*1024*1024*1024, 'pods': 10},
+            'ready': False,
+            'unschedulable': False,
+            'master': True
+         }}
+
 
 def test_get_kube_api(monkeypatch):
     kube_config = MagicMock()
@@ -337,7 +365,7 @@ def test_main(monkeypatch):
     monkeypatch.setattr('kube_aws_autoscaler.main.autoscale', autoscale)
     monkeypatch.setattr('sys.argv', ['foo', '--once', '--dry-run'])
     main()
-    autoscale.assert_called_once_with({'memory': 10, 'pods': 10, 'cpu': 10}, {'memory': 209715200, 'pods': 10, 'cpu': 0.2}, buffer_spare_nodes=1, dry_run=True)
+    autoscale.assert_called_once_with({'memory': 10, 'pods': 10, 'cpu': 10}, {'memory': 209715200, 'pods': 10, 'cpu': 0.2}, buffer_spare_nodes=1, include_master_nodes=False, dry_run=True)
 
     autoscale.side_effect = ValueError