Feature: retry on subset of jobs hosts

awx/main/models/jobs.py

@@ -23,6 +23,9 @@ from django.utils.timezone import utc
 from django.utils.translation import ugettext_lazy as _
 from django.core.exceptions import ValidationError
 
+# REST Framework
+from rest_framework.exceptions import ParseError
+
 # AWX
 from awx.api.versioning import reverse
 from awx.main.models.base import * # noqa
@@ -588,10 +591,33 @@ class Job(UnifiedJob, JobOptions, SurveyJobMixin, JobNotificationMixin, TaskMana
         return self.passwords_needed_to_start
 
     def _get_hosts(self, **kwargs):
-        from awx.main.models.inventory import Host
+        Host = JobHostSummary._meta.get_field('host').related_model
         kwargs['job_host_summaries__job__pk'] = self.pk
         return Host.objects.filter(**kwargs)
 
+    def retry_qs(self, status):
+        '''
+        Returns Host queryset that will be used to produce the `limit`
+        field in a retry on a subset of hosts
+        '''
+        kwargs = {}
+        if status == 'all':
+            pass
+        elif status == 'failed':
+            # Special case for parity with Ansible .retry files
+            kwargs['job_host_summaries__failed'] = True
+        elif status in ['ok', 'changed', 'unreachable']:
+            if status == 'unreachable':
+                status_field = 'dark'
+            else:
+                status_field = status
+            kwargs['job_host_summaries__{}__gt'.format(status_field)] = 0
+        else:
+            raise ParseError(_(
+                '{status_value} is not a valid status option.'
+            ).format(status_value=status))
+        return self._get_hosts(**kwargs)
+
     @property
     def task_impact(self):
         # NOTE: We sorta have to assume the host count matches and that forks default to 5

awx/main/models/unified_jobs.py

@@ -734,7 +734,7 @@ class UnifiedJob(PolymorphicModel, PasswordFieldsModel, CommonModelNameNotUnique
                 pass
         super(UnifiedJob, self).delete()
 
-    def copy_unified_job(self):
+    def copy_unified_job(self, limit=None):
         '''
         Returns saved object, including related fields.
         Create a copy of this unified job for the purpose of relaunch
@@ -746,6 +746,8 @@ class UnifiedJob(PolymorphicModel, PasswordFieldsModel, CommonModelNameNotUnique
         fields = unified_jt_class._get_unified_job_field_names() + [parent_field_name]
         unified_job = copy_model_by_class(self, unified_job_class, fields, {})
         unified_job.launch_type = 'relaunch'
+        if limit:
+            unified_job.limit = limit
         unified_job.save()
 
         # Labels coppied here

awx/main/tests/functional/api/test_job.py

@@ -37,3 +37,31 @@ def test_job_relaunch_permission_denied_response(
     r = post(reverse('api:job_relaunch', kwargs={'pk':job.pk}), {}, jt_user, expect=403)
     assert 'launched with prompted fields' in r.data['detail']
     assert 'do not have permission' in r.data['detail']
+
+
+@pytest.mark.django_db
+@pytest.mark.parametrize("status,hosts", [
+    ('all', 'host1,host2,host3'),
+    ('failed', 'host3'),
+])
+def test_job_relaunch_on_failed_hosts(post, inventory, project, machine_credential, admin_user, status, hosts):
+    h1 = inventory.hosts.create(name='host1')  # no-op
+    h2 = inventory.hosts.create(name='host2')  # changed host
+    h3 = inventory.hosts.create(name='host3')  # failed host
+    jt = JobTemplate.objects.create(
+        name='testjt', inventory=inventory,
+        project=project, credential=machine_credential
+    )
+    job = jt.create_unified_job(_eager_fields={'status': 'failed', 'limit': 'host1,host2,host3'})
+    job.job_events.create(event='playbook_on_stats')
+    job.job_host_summaries.create(host=h1, failed=False, ok=1, changed=0, failures=0, host_name=h1.name)
+    job.job_host_summaries.create(host=h2, failed=False, ok=0, changed=1, failures=0, host_name=h2.name)
+    job.job_host_summaries.create(host=h3, failed=False, ok=0, changed=0, failures=1, host_name=h3.name)
+
+    r = post(
+        url=reverse('api:job_relaunch', kwargs={'pk':job.pk}),
+        data={'hosts': status},
+        user=admin_user,
+        expect=201
+    )
+    assert r.data.get('limit') == hosts