Examples with Job com.spotify.helios.common.descriptors.Job used on opensource projects

Example 56 with Job

use of com.spotify.helios.common.descriptors.Job in project helios by spotify.

the class DeploymentTest method testLotsOfConcurrentJobs.

@Test
public void testLotsOfConcurrentJobs() throws Exception {
    startDefaultMaster();
    final HeliosClient client = defaultClient();
    startDefaultAgent(testHost());
    awaitHostRegistered(client, testHost(), LONG_WAIT_SECONDS, SECONDS);
    awaitHostStatus(client, testHost(), UP, LONG_WAIT_SECONDS, SECONDS);
    final int numberOfJobs = 40;
    final List<JobId> jobIds = Lists.newArrayListWithCapacity(numberOfJobs);
    final String jobName = testJobName + "_" + toHexString(ThreadLocalRandom.current().nextInt());
    // create and deploy a bunch of jobs
    for (Integer i = 0; i < numberOfJobs; i++) {
        final Job job = Job.newBuilder().setName(jobName).setVersion(i.toString()).setImage(BUSYBOX).setCommand(IDLE_COMMAND).setCreatingUser(TEST_USER).build();
        final JobId jobId = job.getId();
        final CreateJobResponse created = client.createJob(job).get();
        assertEquals(CreateJobResponse.Status.OK, created.getStatus());
        final Deployment deployment = Deployment.of(jobId, START, TEST_USER);
        final JobDeployResponse deployed = client.deploy(deployment, testHost()).get();
        assertEquals(JobDeployResponse.Status.OK, deployed.getStatus());
        jobIds.add(jobId);
    }
    // get the container ID's for the jobs
    final Set<String> containerIds = Sets.newHashSetWithExpectedSize(numberOfJobs);
    for (final JobId jobId : jobIds) {
        final TaskStatus taskStatus = awaitJobState(client, testHost(), jobId, RUNNING, LONG_WAIT_SECONDS, SECONDS);
        containerIds.add(taskStatus.getContainerId());
    }
    try (final DockerClient dockerClient = getNewDockerClient()) {
        // kill all the containers for the jobs
        for (final String containerId : containerIds) {
            dockerClient.killContainer(containerId);
        }
        // make sure all the containers come back up
        final int restartedContainers = Polling.await(LONG_WAIT_SECONDS, SECONDS, new Callable<Integer>() {

            @Override
            public Integer call() throws Exception {
                int matchingContainerCount = 0;
                for (final Container c : dockerClient.listContainers()) {
                    for (final String name : c.names()) {
                        if (name.contains(jobName)) {
                            matchingContainerCount++;
                        }
                    }
                }
                if (matchingContainerCount < containerIds.size()) {
                    return null;
                } else {
                    return matchingContainerCount;
                }
            }
        });
        assertEquals(numberOfJobs, restartedContainers);
    }
}

Example 57 with Job

use of com.spotify.helios.common.descriptors.Job in project helios by spotify.

the class DeregisterTest method testDeregisterJobDeployedWithoutStatus.

// Verify that we can deregister a host there are jobs deployed to it, for which there's no
// corresponding status information. For example, if a job was deployed to the host after is went
// down.
@Test
public void testDeregisterJobDeployedWithoutStatus() throws Exception {
    startDefaultMaster();
    final String host = testHost();
    final HeliosClient client = defaultClient();
    final DefaultZooKeeperClient zkClient = new DefaultZooKeeperClient(zk().curatorWithSuperAuth());
    final String idPath = Paths.configHostId(host);
    ZooKeeperRegistrarUtil.registerHost(zkClient, idPath, host, UUID.randomUUID().toString());
    // Create a job
    final Job job = Job.newBuilder().setName(testJobName).setVersion(testJobVersion).setImage(BUSYBOX).setCommand(IDLE_COMMAND).setPorts(ImmutableMap.of("foo", PortMapping.of(4711), "bar", PortMapping.of(4712, ports.localPort("bar")))).build();
    final JobId jobId = job.getId();
    final CreateJobResponse created = client.createJob(job).get();
    assertEquals(CreateJobResponse.Status.OK, created.getStatus());
    // Deploy the job on the agent
    final Deployment deployment = Deployment.of(jobId, START);
    final JobDeployResponse deployed = client.deploy(deployment, host).get();
    assertEquals(JobDeployResponse.Status.OK, deployed.getStatus());
    // Deregister agent
    final HostDeregisterResponse deregisterResponse = client.deregisterHost(host).get();
    assertEquals(HostDeregisterResponse.Status.OK, deregisterResponse.getStatus());
    // Verify that it's possible to remove the job
    final JobDeleteResponse deleteResponse = client.deleteJob(jobId).get();
    assertEquals(JobDeleteResponse.Status.OK, deleteResponse.getStatus());
}

Example 58 with Job

use of com.spotify.helios.common.descriptors.Job in project helios by spotify.

the class FlappingTest method test.

@Test
public void test() throws Exception {
    // CircleCI boxes are too slow -- the job doesn't stop or restart fast enough to ever flap
    assumeFalse(isCircleCi());
    startDefaultMaster();
    final String host = testHost();
    startDefaultAgent(host);
    final HeliosClient client = defaultClient();
    awaitHostStatus(client, host, UP, LONG_WAIT_SECONDS, SECONDS);
    final Job flapper = Job.newBuilder().setName(testJobName).setVersion(testJobVersion).setImage(BUSYBOX).setCommand(asList("nc", "-p", "4711", "-l")).addPort("poke", PortMapping.of(4711)).build();
    final JobId jobId = createJob(flapper);
    deployJob(jobId, host);
    awaitTaskState(jobId, host, RUNNING);
    // Poke the container to make it exit until it's classified as flapping
    Polling.await(LONG_WAIT_SECONDS, SECONDS, new Callable<Object>() {

        @Override
        public Object call() throws Exception {
            final JobStatus jobStatus = getOrNull(client.jobStatus(jobId));
            final TaskStatus taskStatus = jobStatus.getTaskStatuses().get(host);
            if (taskStatus.getThrottled() == FLAPPING) {
                return true;
            }
            final PortMapping port = taskStatus.getPorts().get("poke");
            assert port.getExternalPort() != null;
            poke(port.getExternalPort());
            return null;
        }
    });
    // Verify that the job recovers after we stop poking
    awaitJobThrottle(client, host, jobId, NO, LONG_WAIT_SECONDS, SECONDS);
}

Example 59 with Job

use of com.spotify.helios.common.descriptors.Job in project helios by spotify.

the class HealthCheckTest method testContainerDiesDuringHealthcheck.

@Test
public void testContainerDiesDuringHealthcheck() throws Exception {
    startDefaultMaster();
    final HeliosClient client = defaultClient();
    startDefaultAgent(testHost(), "--service-registry=" + registryAddress);
    awaitHostStatus(client, testHost(), UP, LONG_WAIT_SECONDS, SECONDS);
    final HealthCheck healthCheck = TcpHealthCheck.of("health");
    final Job job = pokeJob(healthCheck);
    final JobId jobId = createJob(job);
    deployJob(jobId, testHost());
    awaitTaskState(jobId, testHost(), HEALTHCHECKING);
    // kill the underlying container
    final JobStatus jobStatus = getOrNull(client.jobStatus(jobId));
    final TaskStatus taskStatus = jobStatus.getTaskStatuses().get(testHost());
    getNewDockerClient().killContainer(taskStatus.getContainerId());
    // ensure the job is marked as failed
    final int timeout = WAIT_TIMEOUT_SECONDS;
    Polling.await(timeout, SECONDS, new Callable<Object>() {

        @Override
        public Object call() throws Exception {
            final TaskStatusEvents jobHistory = getOrNull(client.jobHistory(jobId));
            for (final TaskStatusEvent event : jobHistory.getEvents()) {
                if (event.getStatus().getState() == FAILED) {
                    return true;
                }
            }
            return null;
        }
    });
    // wait for the job to come back up and start healthchecking again
    awaitTaskState(jobId, testHost(), HEALTHCHECKING);
    pokeAndVerifyRegistration(client, jobId, timeout);
}

Example 60 with Job

use of com.spotify.helios.common.descriptors.Job in project helios by spotify.

the class CliDeploymentTest method testDeployAndUndeployJob.

@Test
public void testDeployAndUndeployJob() throws Exception {
    startDefaultMaster();
    // Wait for master to come up
    Polling.await(LONG_WAIT_SECONDS, SECONDS, new Callable<String>() {

        @Override
        public String call() throws Exception {
            final String output = cli("masters");
            return output.contains(masterName()) ? output : null;
        }
    });
    startDefaultAgent(testHost());
    final String image = BUSYBOX;
    final Map<String, PortMapping> ports = ImmutableMap.of("foo", PortMapping.of(4711), "bar", PortMapping.of(5000, externalPort));
    final Map<ServiceEndpoint, ServicePorts> registration = ImmutableMap.of(ServiceEndpoint.of("foo-service", "tcp"), ServicePorts.of("foo"), ServiceEndpoint.of("bar-service", "http"), ServicePorts.of("bar"));
    final Map<String, String> env = ImmutableMap.of("BAD", "f00d");
    // Wait for agent to come up
    awaitHostRegistered(testHost(), LONG_WAIT_SECONDS, SECONDS);
    awaitHostStatus(testHost(), UP, LONG_WAIT_SECONDS, SECONDS);
    // Create job
    final JobId jobId = createJob(testJobName, testJobVersion, image, IDLE_COMMAND, env, ports, registration);
    // Query for job
    final Job expected = Job.newBuilder().setName(testJobName).setVersion(testJobVersion).setImage(image).setCommand(IDLE_COMMAND).setEnv(env).setPorts(ports).setRegistration(registration).setCreatingUser(TEST_USER).build();
    final String inspectOutput = cli("inspect", "--json", jobId.toString());
    final Job parsed = Json.read(inspectOutput, Job.class);
    assertJobEquals(expected, parsed);
    assertThat(cli("jobs", testJobName, "-q"), containsString(jobId.toString()));
    assertThat(cli("jobs", testJobName + ":" + testJobVersion, "-q"), containsString(jobId.toString()));
    assertEquals("job pattern foozbarz matched no jobs", cli("jobs", "foozbarz").trim());
    assertTrue(cli("jobs", "foozbarz", "-q").isEmpty());
    // Create a new job using the first job as a template
    final Job expectedCloned = expected.toBuilder().setVersion(expected.getId().getVersion() + "-cloned").build();
    final JobId clonedJobId = JobId.parse(WHITESPACE.trimFrom(cli("create", "-q", "-t", testJobName + ":" + testJobVersion, testJobName + ":" + testJobVersion + "-cloned")));
    final String clonedInspectOutput = cli("inspect", "--json", clonedJobId.toString());
    final Job clonedParsed = Json.read(clonedInspectOutput, Job.class);
    assertJobEquals(expectedCloned, clonedParsed);
    // Verify that port mapping and environment variables are correct
    final String statusString = cli("status", "--job", jobId.toString(), "--json");
    final Map<JobId, JobStatus> statuses = Json.read(statusString, STATUSES_TYPE);
    final Job job = statuses.get(jobId).getJob();
    assertEquals(ServicePorts.of("foo"), job.getRegistration().get(ServiceEndpoint.of("foo-service", "tcp")));
    assertEquals(ServicePorts.of("bar"), job.getRegistration().get(ServiceEndpoint.of("bar-service", "http")));
    assertEquals(4711, job.getPorts().get("foo").getInternalPort());
    assertEquals(PortMapping.of(5000, externalPort), job.getPorts().get("bar"));
    assertEquals("f00d", job.getEnv().get("BAD"));
    final String duplicateJob = cli("create", testJobName + ":" + testJobVersion, image, "--", IDLE_COMMAND);
    assertThat(duplicateJob, containsString("JOB_ALREADY_EXISTS"));
    final String prestop = stopJob(jobId, testHost());
    assertThat(prestop, containsString("JOB_NOT_DEPLOYED"));
    // Deploy job
    deployJob(jobId, testHost());
    // Stop job
    final String stop1 = stopJob(jobId, BOGUS_HOST);
    assertThat(stop1, containsString("HOST_NOT_FOUND"));
    final String stop2 = stopJob(BOGUS_JOB, testHost());
    assertThat(stop2, containsString("Unknown job"));
    final String stop3 = stopJob(jobId, testHost());
    assertThat(stop3, containsString(testHost() + ": done"));
    // Verify that undeploying the job from a nonexistent host fails
    assertThat(cli("undeploy", jobId.toString(), BOGUS_HOST), containsString("HOST_NOT_FOUND"));
    // Verify that undeploying a nonexistent job from the host fails
    assertThat(cli("undeploy", BOGUS_JOB.toString(), testHost()), containsString("Unknown job"));
    // Undeploy job
    undeployJob(jobId, testHost());
}