use of com.spotify.helios.common.descriptors.TaskStatus in project helios by spotify.
the class TerminationTest method testTermOnExit.
@Test
public void testTermOnExit() throws Exception {
startDefaultMaster();
final String host = testHost();
startDefaultAgent(host);
final HeliosClient client = defaultClient();
awaitHostStatus(client, host, UP, LONG_WAIT_SECONDS, SECONDS);
// Note: signal 15 is SIGTERM
final Job jobToInterrupt = Job.newBuilder().setName(testJobName).setVersion(testJobVersion).setImage(BUSYBOX).setCommand(asList("/bin/sh", "-c", "trap handle 15; handle() { echo term; exit 0; }; " + "while true; do sleep 1; done")).build();
final JobId jobId = createJob(jobToInterrupt);
deployJob(jobId, host);
awaitTaskState(jobId, host, RUNNING);
client.setGoal(new Deployment(jobId, Goal.STOP, Deployment.EMTPY_DEPLOYER_USER, Deployment.EMPTY_DEPLOYER_MASTER, Deployment.EMPTY_DEPLOYMENT_GROUP_NAME), host);
final TaskStatus taskStatus = awaitTaskState(jobId, host, STOPPED);
final String log;
try (final DockerClient dockerClient = getNewDockerClient();
LogStream logs = dockerClient.logs(taskStatus.getContainerId(), stdout())) {
log = logs.readFully();
}
// Message expected, because the SIGTERM handler in the script should have run
assertEquals("term\n", log);
}
use of com.spotify.helios.common.descriptors.TaskStatus 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);
}
}
use of com.spotify.helios.common.descriptors.TaskStatus in project helios by spotify.
the class DnsServerTest method testDnsParam.
@Test
public void testDnsParam() throws Exception {
final String server1 = "127.0.0.1";
final String server2 = "127.0.0.2";
startDefaultMaster();
startDefaultAgent(testHost(), "--dns", server1, "--dns", server2);
awaitHostStatus(testHost(), UP, LONG_WAIT_SECONDS, SECONDS);
final JobId jobId = createJob(testJobName, testJobVersion, BUSYBOX, asList("cat", "/etc/resolv.conf"));
deployJob(jobId, testHost());
final TaskStatus taskStatus = awaitTaskState(jobId, testHost(), EXITED);
try (final DockerClient dockerClient = getNewDockerClient()) {
final LogStream logs = dockerClient.logs(taskStatus.getContainerId(), stdout(), stderr());
final String log = logs.readFully();
assertThat(log, containsString(server1));
assertThat(log, containsString(server2));
}
}
use of com.spotify.helios.common.descriptors.TaskStatus 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);
}
use of com.spotify.helios.common.descriptors.TaskStatus 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);
}
Aggregations