Bug 1302227: set the taskGroupId to match the decision task; r=garndt

In this case, the tasks must have the same schedulerId as the existing task,
so this is calculated using parameters['level'].

MozReview-Commit-ID: G8EE2kvFstT

taskcluster/taskgraph/create.py

@@ -23,9 +23,7 @@ logger = logging.getLogger(__name__)
 CONCURRENCY = 50
 
 
-def create_tasks(taskgraph, label_to_taskid):
-    # TODO: use the taskGroupId of the decision task
-    task_group_id = slugid()
+def create_tasks(taskgraph, label_to_taskid, params):
     taskid_to_label = {t: l for l, t in label_to_taskid.iteritems()}
 
     session = requests.Session()
@@ -40,6 +38,13 @@ def create_tasks(taskgraph, label_to_taskid):
 
     decision_task_id = os.environ.get('TASK_ID')
 
+    # when running as an actual decision task, we use the decision task's
+    # taskId as the taskGroupId.  The process that created the decision task
+    # helpfully placed it in this same taskGroup.  If there is no $TASK_ID,
+    # fall back to a slugid
+    task_group_id = decision_task_id or slugid()
+    scheduler_id = 'gecko-level-{}'.format(params['level'])
+
     with futures.ThreadPoolExecutor(CONCURRENCY) as e:
         fs = {}
 
@@ -62,7 +67,7 @@ def create_tasks(taskgraph, label_to_taskid):
                 task_def['dependencies'] = [decision_task_id]
 
             task_def['taskGroupId'] = task_group_id
-            task_def['schedulerId'] = '-'
+            task_def['schedulerId'] = scheduler_id
 
             # Wait for dependencies before submitting this.
             deps_fs = [fs[dep] for dep in task_def.get('dependencies', [])