Calling start on running container doesn't restart

[#150536302]

Signed-off-by: Joe Kirwin <jkirwin@pivotal.io>

Author: Anoop Gopalakrishnan

src/bpm/commands/start.go

@@ -17,6 +17,7 @@ package commands
 
 import (
 	"bpm/config"
+	"bpm/runc/lifecycle"
 	"fmt"
 	"path/filepath"
 
@@ -74,23 +75,34 @@ func start(cmd *cobra.Command, _ []string) error {
 	}
 
 	runcLifecycle := newRuncLifecycle()
+	state := getContainerState(runcLifecycle, bpmCfg.JobName())
 
-	err = runcLifecycle.StartJob(bpmCfg, procCfg)
-	if err != nil {
+	switch state {
+	case lifecycle.ContainerStateRunning:
+		logger.Info(fmt.Sprintf("container %s is already running", bpmCfg.JobName()))
+		return nil
+	case lifecycle.ContainerStateStopped:
 		removeErr := runcLifecycle.RemoveJob(bpmCfg)
 		if removeErr != nil {
 			logger.Error("failed-to-cleanup", removeErr)
 			return removeErr
 		}
+		return runcLifecycle.StartJob(bpmCfg, procCfg)
+	default:
+		return runcLifecycle.StartJob(bpmCfg, procCfg)
+	}
+}
 
-		// Case of pre-existing stopped container
-		// with same name that needed to be cleaned up
-		err = runcLifecycle.StartJob(bpmCfg, procCfg)
+func getContainerState(runcLifecycle *lifecycle.RuncLifecycle, jobName string) string {
+	jobs, err := runcLifecycle.ListJobs()
+	if err != nil {
+		logger.Error("failed-list-jobs", err)
+	}
 
-		if err != nil {
-			return err
+	for _, job := range jobs {
+		if job.Name == jobName {
+			return job.Status
 		}
 	}
-
-	return nil
+	return ""
 }

src/bpm/main_test.go

@@ -613,6 +613,64 @@ var _ = Describe("bpm", func() {
 			})
 		})
 
+		Context("when a running container exist with the same name", func() {
+			startContainer := func() *exec.Cmd {
+				cfg.Executable = "/bin/bash"
+				cfg.Args = []string{
+					"-c",
+					"sleep 10000",
+				}
+
+				limit := int64(1000)
+				cfg.Limits = &config.Limits{
+					Processes: &limit,
+				}
+
+				cfgPath = writeConfig(jobName, jobName, cfg)
+
+				start := exec.Command(bpmPath, "start", jobName)
+				start.Env = append(start.Env, fmt.Sprintf("BPM_BOSH_ROOT=%s", boshConfigPath))
+				return start
+			}
+
+			// Currently assumes only one thing is running #NAIVE
+			getContainerPid := func() int {
+				command = exec.Command(bpmPath, "list")
+				command.Env = append(command.Env, fmt.Sprintf("BPM_BOSH_ROOT=%s", boshConfigPath))
+				session, err := gexec.Start(command, GinkgoWriter, GinkgoWriter)
+				Expect(err).ShouldNot(HaveOccurred())
+				Eventually(session).Should(gexec.Exit(0))
+				Expect(session.Out).Should(gbytes.Say("running"))
+
+				re := regexp.MustCompile("\\s(\\d+)\\s")
+				pids := re.FindSubmatch(session.Out.Contents())
+				Expect(pids).ShouldNot(BeNil())
+				Expect(len(pids)).Should(Equal(2))
+
+				pid, err := strconv.Atoi(string(pids[1]))
+				Expect(err).NotTo(HaveOccurred())
+				return pid
+			}
+
+			BeforeEach(func() {
+				start := startContainer()
+				session, err := gexec.Start(start, GinkgoWriter, GinkgoWriter)
+				Expect(err).ShouldNot(HaveOccurred())
+				Eventually(session).Should(gexec.Exit(0))
+			})
+
+			It("should not restart the container and logs", func() {
+				origPid := getContainerPid()
+				start := startContainer()
+				session, err := gexec.Start(start, GinkgoWriter, GinkgoWriter)
+				Expect(err).ShouldNot(HaveOccurred())
+				Eventually(session).Should(gexec.Exit(0))
+				Expect(session.Out).Should(gbytes.Say(fmt.Sprintf("container %s is already running", jobName)))
+				newPid := getContainerPid()
+				Expect(newPid).To(Equal(origPid))
+			})
+		})
+
 		Context("when a stopped container exists with the same name", func() {
 			BeforeEach(func() {
 				cfg.Executable = "/bin/bash"

src/bpm/runc/lifecycle/lifecycle.go

@@ -36,6 +36,9 @@ import (
 const (
 	ContainerSigQuitGracePeriod = 5 * time.Second
 	ContainerStatePollInterval  = 1 * time.Second
+	ContainerStateRunning       = "running"
+	ContainerStatePaused        = "paused"
+	ContainerStateStopped       = "stopped"
 )
 
 var TimeoutError = errors.New("failed to stop job within timeout")