Adding parameter for Aurora so that we're able to run SLA aware updates with less than 20 instances. Lowered time it takes to run test by reducing watch time per instance as well.

docker-compose.yml

@@ -55,6 +55,30 @@ services:
     depends_on:
     - zk
 
+  agent-two:
+    image: rdelvalle/mesos-agent:1.5.1
+    pid: host
+    restart: on-failure
+    ports:
+      - "5061:5061"
+    environment:
+      MESOS_MASTER: zk://192.168.33.2:2181/mesos
+      MESOS_CONTAINERIZERS: docker,mesos
+      MESOS_HOSTNAME: localhost
+      MESOS_PORT: 5061
+      MESOS_RESOURCES: ports(*):[11000-11999]
+      MESOS_SYSTEMD_ENABLE_SUPPORT: 'false'
+      MESOS_WORK_DIR: /tmp/mesos
+    networks:
+      aurora_cluster:
+        ipv4_address: 192.168.33.5
+
+    volumes:
+      - /sys/fs/cgroup:/sys/fs/cgroup
+      - /var/run/docker.sock:/var/run/docker.sock
+    depends_on:
+      - zk
+
   aurora-one:
     image: rdelvalle/aurora:0.21.0
     pid: host
@@ -65,6 +89,7 @@ services:
       CLUSTER_NAME: test-cluster
       ZK_ENDPOINTS: "192.168.33.2:2181"
       MESOS_MASTER: "zk://192.168.33.2:2181/mesos"
+      EXTRA_SCHEDULER_ARGS: "-min_required_instances_for_sla_check=1"
     networks:
       aurora_cluster:
         ipv4_address: 192.168.33.7

job.go

@@ -56,6 +56,8 @@ type Job interface {
 	MaxFailure(maxFail int32) Job
 	Container(container Container) Job
 	PartitionPolicy(policy *aurora.PartitionPolicy) Job
+	Tier(tier string) Job
+	SlaPolicy(policy *aurora.SlaPolicy) Job
 }
 
 type ResourceType int
@@ -323,6 +325,19 @@ func (j *AuroraJob) Container(container Container) Job {
 // Set a partition policy for the job configuration to implement.
 func (j *AuroraJob) PartitionPolicy(policy *aurora.PartitionPolicy) Job {
 	j.jobConfig.TaskConfig.PartitionPolicy = policy
+	return j
+}
+
+// Set the Tier for the Job.
+func (j *AuroraJob) Tier(tier string) Job {
+	j.jobConfig.TaskConfig.Tier = &tier
+
+	return j
+}
+
+// Set an SlaPolicy for the Job.
+func (j *AuroraJob) SlaPolicy(policy *aurora.SlaPolicy) Job {
+	j.jobConfig.TaskConfig.SlaPolicy = policy
 
 	return j
 }

realis_e2e_test.go

@@ -750,3 +750,76 @@ func TestRealisClient_PartitionPolicy(t *testing.T) {
 	// Clean up after finishing test
 	_, err = r.KillJob(job.JobKey())
 }
+
+func TestAuroraJob_UpdateSlaPolicy(t *testing.T) {
+
+	tests := []struct {
+		name string
+		args aurora.SlaPolicy
+	}{
+		{
+			"create_service_with_sla_count_policy_test",
+			aurora.SlaPolicy{CountSlaPolicy: &aurora.CountSlaPolicy{Count: 1, DurationSecs: 30}},
+		},
+		{
+			"create_service_with_sla_percentage_policy_test",
+			aurora.SlaPolicy{PercentageSlaPolicy: &aurora.PercentageSlaPolicy{Percentage: 0.25, DurationSecs: 30}},
+		},
+		{
+			"create_service_with_sla_coordinator_policy_test",
+			aurora.SlaPolicy{CoordinatorSlaPolicy: &aurora.CoordinatorSlaPolicy{
+				CoordinatorUrl: "http://localhost/endpoint", StatusKey: "aurora_test"}},
+		},
+	}
+	role := "vagrant"
+
+	_, err := r.SetQuota(role, thrift.Float64Ptr(6.0), thrift.Int64Ptr(1024), thrift.Int64Ptr(1024))
+	assert.NoError(t, err)
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+
+			// Create a single job
+			job := realis.NewJob().
+				Environment("prod").
+				Role(role).
+				Name(tt.name).
+				ExecutorName(aurora.AURORA_EXECUTOR_NAME).
+				ExecutorData(string(thermosPayload)).
+				CPU(.01).
+				RAM(2).
+				Disk(5).
+				InstanceCount(8).
+				IsService(true).
+				SlaPolicy(&tt.args).
+				Tier("preferred")
+
+			settings := realis.NewUpdateSettings()
+			settings.UpdateGroupSize = 4
+			settings.MinWaitInInstanceRunningMs = 5 * 1000
+
+			_, result, err := r.CreateService(job, settings)
+			assert.NoError(t, err)
+			assert.NotNil(t, result)
+
+			var ok bool
+			var mErr error
+
+			if ok, mErr = monitor.JobUpdate(*result.GetKey(), 5, 240); !ok || mErr != nil {
+				// Update may already be in a terminal state so don't check for error
+				_, err := r.AbortJobUpdate(*result.GetKey(), "Monitor timed out.")
+
+				_, err = r.KillJob(job.JobKey())
+
+				assert.NoError(t, err)
+			}
+			assert.True(t, ok)
+			assert.NoError(t, mErr)
+
+			// Kill task test task after confirming it came up fine
+			_, err = r.KillJob(job.JobKey())
+
+			assert.NoError(t, err)
+		})
+	}
+}

runTestsMac.sh

@@ -1,4 +1,4 @@
 #!/bin/bash
 
 # Since we run our docker compose setup in bridge mode to be able to run on MacOS, we have to launch a Docker container within the bridge network in order to avoid any routing issues.
-docker run -t -v $(pwd):/go/src/github.com/paypal/gorealis --network gorealis_aurora_cluster golang:1.10.3-stretch  go test -v github.com/paypal/gorealis $@
+docker run -t -v $(pwd):/go/src/github.com/paypal/gorealis --network gorealis_aurora_cluster golang:1.10-stretch go test -v github.com/paypal/gorealis $@