@Test
public void testMultipleResubmits() throws Exception {
LOG.info("TestMultipleResbmits - no indefinite resubmissions");
conf.setInt("hbase.splitlog.max.resubmit", 2);
slm = new SplitLogManager(zkw, conf, stopper, master, DUMMY_MASTER, null);
slm.finishInitialization();
TaskBatch batch = new TaskBatch();
String tasknode = submitTaskAndWait(batch, "foo/1");
int version = ZKUtil.checkExists(zkw, tasknode);
final ServerName worker1 = new ServerName("worker1,1,1");
final ServerName worker2 = new ServerName("worker2,1,1");
final ServerName worker3 = new ServerName("worker3,1,1");
SplitLogTask slt = new SplitLogTask.Owned(worker1);
ZKUtil.setData(zkw, tasknode, slt.toByteArray());
waitForCounter(tot_mgr_heartbeat, 0, 1, to / 2);
waitForCounter(tot_mgr_resubmit, 0, 1, to + to / 2);
int version1 = ZKUtil.checkExists(zkw, tasknode);
assertTrue(version1 > version);
slt = new SplitLogTask.Owned(worker2);
ZKUtil.setData(zkw, tasknode, slt.toByteArray());
waitForCounter(tot_mgr_heartbeat, 1, 2, to / 2);
waitForCounter(tot_mgr_resubmit, 1, 2, to + to / 2);
int version2 = ZKUtil.checkExists(zkw, tasknode);
assertTrue(version2 > version1);
slt = new SplitLogTask.Owned(worker3);
ZKUtil.setData(zkw, tasknode, slt.toByteArray());
waitForCounter(tot_mgr_heartbeat, 1, 2, to / 2);
waitForCounter(tot_mgr_resubmit_threshold_reached, 0, 1, to + to / 2);
Thread.sleep(to + to / 2);
assertEquals(2L, tot_mgr_resubmit.get());
}
@Test
public void testMultipleResubmits() throws Exception {
LOG.info("TestMultipleResbmits - no indefinite resubmissions");
int to = 1000;
conf.setInt("hbase.splitlog.manager.timeout", to);
conf.setInt("hbase.splitlog.manager.timeoutmonitor.period", 100);
to = to + 2 * 100;
conf.setInt("hbase.splitlog.max.resubmit", 2);
slm = new SplitLogManager(zkw, conf, stopper, "dummy-master", null);
slm.finishInitialization();
TaskBatch batch = new TaskBatch();
String tasknode = submitTaskAndWait(batch, "foo/1");
int version = ZKUtil.checkExists(zkw, tasknode);
ZKUtil.setData(zkw, tasknode, TaskState.TASK_OWNED.get("worker1"));
waitForCounter(tot_mgr_heartbeat, 0, 1, 1000);
waitForCounter(tot_mgr_resubmit, 0, 1, to + 100);
int version1 = ZKUtil.checkExists(zkw, tasknode);
assertTrue(version1 > version);
ZKUtil.setData(zkw, tasknode, TaskState.TASK_OWNED.get("worker2"));
waitForCounter(tot_mgr_heartbeat, 1, 2, 1000);
waitForCounter(tot_mgr_resubmit, 1, 2, to + 100);
int version2 = ZKUtil.checkExists(zkw, tasknode);
assertTrue(version2 > version1);
ZKUtil.setData(zkw, tasknode, TaskState.TASK_OWNED.get("worker3"));
waitForCounter(tot_mgr_heartbeat, 1, 2, 1000);
waitForCounter(tot_mgr_resubmit_threshold_reached, 0, 1, to + 100);
Thread.sleep(to + 100);
assertEquals(2L, tot_mgr_resubmit.get());
}
public void updateRecord(MetaRecord rec) {
try {
long ts = System.currentTimeMillis();
putRecord(rec, ts);
String node = ZKUtil.joinZNode(H2MetaTableTracker.NODE_NAME, Integer.toString(rec.getId()));
// setData会异步触发所有机器(包括本机)上的H2MetaTableTracker.nodeDataChanged
// 然后触发下列调用:
// =>org.h2.engine.Database.updateDatabaseObject(int)
// =>org.h2.engine.Database.update(Session, DbObject)
// =>org.h2.engine.Database.addMeta0(Session, DbObject, boolean)
// =>又到此方法
// 所以会造成循环
synchronized (this) { // 避免setData后立刻触发nodeDataChanged，此时IdVersion还未更新
ZKUtil.setData(watcher, node, Bytes.toBytes(ts));
// setData后watch不见了，所以要继续watch，监听其他人对此node的修改
// ZKUtil.watchAndCheckExists(watcher, node);
Stat stat = new Stat();
ZKUtil.getDataAndWatch(watcher, node, stat);
// 这里记录下id的最新版本，触发nodeDataChanged时再检查一下是否版本一样，
// 如果不大于这里的版本那么就不再执行updateDatabaseObject操作
tracker.updateIdVersion(rec.getId(), stat.getVersion());
}
} catch (Exception e) {
throw new RuntimeException(e);
}
}
@Test
public void testRescanCleanup() throws Exception {
LOG.info("TestRescanCleanup - ensure RESCAN nodes are cleaned up");
slm = new SplitLogManager(zkw, conf, stopper, master, DUMMY_MASTER, null);
slm.finishInitialization();
TaskBatch batch = new TaskBatch();
String tasknode = submitTaskAndWait(batch, "foo/1");
int version = ZKUtil.checkExists(zkw, tasknode);
final ServerName worker1 = new ServerName("worker1,1,1");
SplitLogTask slt = new SplitLogTask.Owned(worker1);
ZKUtil.setData(zkw, tasknode, slt.toByteArray());
waitForCounter(tot_mgr_heartbeat, 0, 1, to / 2);
waitForCounter(
new Expr() {
@Override
public long eval() {
return (tot_mgr_resubmit.get() + tot_mgr_resubmit_failed.get());
}
},
0,
1,
5 * 60000); // wait long enough
Assert.assertEquals(
"Could not run test. Lost ZK connection?", 0, tot_mgr_resubmit_failed.get());
int version1 = ZKUtil.checkExists(zkw, tasknode);
assertTrue(version1 > version);
byte[] taskstate = ZKUtil.getData(zkw, tasknode);
slt = SplitLogTask.parseFrom(taskstate);
assertTrue(slt.isUnassigned(DUMMY_MASTER));
waitForCounter(tot_mgr_rescan_deleted, 0, 1, to / 2);
}
@Test
public void testDeadWorker() throws Exception {
LOG.info("testDeadWorker");
conf.setLong("hbase.splitlog.max.resubmit", 0);
slm = new SplitLogManager(zkw, conf, stopper, master, DUMMY_MASTER, null);
slm.finishInitialization();
TaskBatch batch = new TaskBatch();
String tasknode = submitTaskAndWait(batch, "foo/1");
int version = ZKUtil.checkExists(zkw, tasknode);
final ServerName worker1 = new ServerName("worker1,1,1");
SplitLogTask slt = new SplitLogTask.Owned(worker1);
ZKUtil.setData(zkw, tasknode, slt.toByteArray());
if (tot_mgr_heartbeat.get() == 0) waitForCounter(tot_mgr_heartbeat, 0, 1, to / 2);
slm.handleDeadWorker(worker1);
if (tot_mgr_resubmit.get() == 0) waitForCounter(tot_mgr_resubmit, 0, 1, to + to / 2);
if (tot_mgr_resubmit_dead_server_task.get() == 0) {
waitForCounter(tot_mgr_resubmit_dead_server_task, 0, 1, to + to / 2);
}
int version1 = ZKUtil.checkExists(zkw, tasknode);
assertTrue(version1 > version);
byte[] taskstate = ZKUtil.getData(zkw, tasknode);
slt = SplitLogTask.parseFrom(taskstate);
assertTrue(slt.isUnassigned(DUMMY_MASTER));
return;
}
/**
* endTask() can fail and the only way to recover out of it is for the {@link SplitLogManager} to
* timeout the task node.
*
* @param slt
* @param ctr
*/
public static void endTask(
ZooKeeperWatcher zkw, SplitLogTask slt, AtomicLong ctr, String task, int taskZKVersion) {
try {
if (ZKUtil.setData(zkw, task, slt.toByteArray(), taskZKVersion)) {
LOG.info("successfully transitioned task " + task + " to final state " + slt);
ctr.incrementAndGet();
return;
}
LOG.warn(
"failed to transistion task "
+ task
+ " to end state "
+ slt
+ " because of version mismatch ");
} catch (KeeperException.BadVersionException bve) {
LOG.warn(
"transisition task " + task + " to " + slt + " failed because of version mismatch", bve);
} catch (KeeperException.NoNodeException e) {
LOG.fatal(
"logic error - end task " + task + " " + slt + " failed because task doesn't exist", e);
} catch (KeeperException e) {
LOG.warn("failed to end task, " + task + " " + slt, e);
}
SplitLogCounters.tot_wkr_final_transition_failed.incrementAndGet();
}
/**
* endTask() can fail and the only way to recover out of it is for the {@link
* org.apache.hadoop.hbase.master.SplitLogManager} to timeout the task node.
*
* @param slt
* @param ctr
*/
@Override
public void endTask(SplitLogTask slt, AtomicLong ctr, SplitTaskDetails details) {
ZkSplitTaskDetails zkDetails = (ZkSplitTaskDetails) details;
String task = zkDetails.getTaskNode();
int taskZKVersion = zkDetails.getCurTaskZKVersion().intValue();
try {
if (ZKUtil.setData(watcher, task, slt.toByteArray(), taskZKVersion)) {
LOG.info("successfully transitioned task " + task + " to final state " + slt);
ctr.incrementAndGet();
return;
}
LOG.warn(
"failed to transistion task "
+ task
+ " to end state "
+ slt
+ " because of version mismatch ");
} catch (KeeperException.BadVersionException bve) {
LOG.warn(
"transisition task " + task + " to " + slt + " failed because of version mismatch", bve);
} catch (KeeperException.NoNodeException e) {
LOG.fatal(
"logic error - end task " + task + " " + slt + " failed because task doesn't exist", e);
} catch (KeeperException e) {
LOG.warn("failed to end task, " + task + " " + slt, e);
}
SplitLogCounters.tot_wkr_final_transition_failed.incrementAndGet();
}
@Test
public void testUnassignedTimeout() throws Exception {
LOG.info("TestUnassignedTimeout - iff all tasks are unassigned then" + " resubmit");
// create an orphan task in OWNED state
String tasknode1 = ZKSplitLog.getEncodedNodeName(zkw, "orphan/1");
final ServerName worker1 = new ServerName("worker1,1,1");
SplitLogTask slt = new SplitLogTask.Owned(worker1);
zkw.getRecoverableZooKeeper()
.create(tasknode1, slt.toByteArray(), Ids.OPEN_ACL_UNSAFE, CreateMode.PERSISTENT);
slm = new SplitLogManager(zkw, conf, stopper, master, DUMMY_MASTER, null);
slm.finishInitialization();
waitForCounter(tot_mgr_orphan_task_acquired, 0, 1, to / 2);
// submit another task which will stay in unassigned mode
TaskBatch batch = new TaskBatch();
submitTaskAndWait(batch, "foo/1");
// keep updating the orphan owned node every to/2 seconds
for (int i = 0; i < (3 * to) / 100; i++) {
Thread.sleep(100);
final ServerName worker2 = new ServerName("worker1,1,1");
slt = new SplitLogTask.Owned(worker2);
ZKUtil.setData(zkw, tasknode1, slt.toByteArray());
}
// since we have stopped heartbeating the owned node therefore it should
// get resubmitted
LOG.info("waiting for manager to resubmit the orphan task");
waitForCounter(tot_mgr_resubmit, 0, 1, to + to / 2);
// now all the nodes are unassigned. manager should post another rescan
waitForCounter(tot_mgr_resubmit_unassigned, 0, 1, 2 * to + to / 2);
}
@Test
public void testTaskResigned() throws Exception {
LOG.info("TestTaskResigned - resubmit task node once in RESIGNED state");
assertEquals(tot_mgr_resubmit.get(), 0);
slm = new SplitLogManager(zkw, conf, stopper, master, DUMMY_MASTER, null);
slm.finishInitialization();
assertEquals(tot_mgr_resubmit.get(), 0);
TaskBatch batch = new TaskBatch();
String tasknode = submitTaskAndWait(batch, "foo/1");
assertEquals(tot_mgr_resubmit.get(), 0);
final ServerName worker1 = new ServerName("worker1,1,1");
assertEquals(tot_mgr_resubmit.get(), 0);
SplitLogTask slt = new SplitLogTask.Resigned(worker1);
assertEquals(tot_mgr_resubmit.get(), 0);
ZKUtil.setData(zkw, tasknode, slt.toByteArray());
int version = ZKUtil.checkExists(zkw, tasknode);
// Could be small race here.
if (tot_mgr_resubmit.get() == 0) waitForCounter(tot_mgr_resubmit, 0, 1, to / 2);
assertEquals(tot_mgr_resubmit.get(), 1);
int version1 = ZKUtil.checkExists(zkw, tasknode);
assertTrue("version1=" + version1 + ", version=" + version, version1 > version);
byte[] taskstate = ZKUtil.getData(zkw, tasknode);
slt = SplitLogTask.parseFrom(taskstate);
assertTrue(slt.isUnassigned(DUMMY_MASTER));
}
@Test
public void testTaskDone() throws Exception {
LOG.info("TestTaskDone - cleanup task node once in DONE state");
slm = new SplitLogManager(zkw, conf, stopper, "dummy-master", null);
slm.finishInitialization();
TaskBatch batch = new TaskBatch();
String tasknode = submitTaskAndWait(batch, "foo/1");
ZKUtil.setData(zkw, tasknode, TaskState.TASK_DONE.get("worker"));
synchronized (batch) {
while (batch.installed != batch.done) {
batch.wait();
}
}
waitForCounter(tot_mgr_task_deleted, 0, 1, 1000);
assertTrue(ZKUtil.checkExists(zkw, tasknode) == -1);
}
@Test
public void testTaskResigned() throws Exception {
LOG.info("TestTaskResigned - resubmit task node once in RESIGNED state");
slm = new SplitLogManager(zkw, conf, stopper, "dummy-master", null);
slm.finishInitialization();
TaskBatch batch = new TaskBatch();
String tasknode = submitTaskAndWait(batch, "foo/1");
ZKUtil.setData(zkw, tasknode, TaskState.TASK_RESIGNED.get("worker"));
int version = ZKUtil.checkExists(zkw, tasknode);
waitForCounter(tot_mgr_resubmit, 0, 1, 1000);
int version1 = ZKUtil.checkExists(zkw, tasknode);
assertTrue(version1 > version);
byte[] taskstate = ZKUtil.getData(zkw, tasknode);
assertTrue(Arrays.equals(taskstate, TaskState.TASK_UNASSIGNED.get("dummy-master")));
}
@Override
public void setPeerTableCFsConfig(String id, String tableCFsStr) throws ReplicationException {
try {
if (!peerExists(id)) {
throw new IllegalArgumentException(
"Cannot set peer tableCFs because id=" + id + " does not exist.");
}
String tableCFsZKNode = getTableCFsNode(id);
byte[] tableCFs = Bytes.toBytes(tableCFsStr);
if (ZKUtil.checkExists(this.zookeeper, tableCFsZKNode) != -1) {
ZKUtil.setData(this.zookeeper, tableCFsZKNode, tableCFs);
} else {
ZKUtil.createAndWatch(this.zookeeper, tableCFsZKNode, tableCFs);
}
LOG.info("Peer tableCFs with id= " + id + " is now " + tableCFsStr);
} catch (KeeperException e) {
throw new ReplicationException("Unable to change tableCFs of the peer with id=" + id, e);
}
}
@Test
public void testTaskDone() throws Exception {
LOG.info("TestTaskDone - cleanup task node once in DONE state");
slm = new SplitLogManager(zkw, conf, stopper, master, DUMMY_MASTER, null);
slm.finishInitialization();
TaskBatch batch = new TaskBatch();
String tasknode = submitTaskAndWait(batch, "foo/1");
final ServerName worker1 = new ServerName("worker1,1,1");
SplitLogTask slt = new SplitLogTask.Done(worker1);
ZKUtil.setData(zkw, tasknode, slt.toByteArray());
synchronized (batch) {
while (batch.installed != batch.done) {
batch.wait();
}
}
waitForCounter(tot_mgr_task_deleted, 0, 1, to / 2);
assertTrue(ZKUtil.checkExists(zkw, tasknode) == -1);
}
@Test
public void testTaskErr() throws Exception {
LOG.info("TestTaskErr - cleanup task node once in ERR state");
conf.setInt("hbase.splitlog.max.resubmit", 0);
slm = new SplitLogManager(zkw, conf, stopper, "dummy-master", null);
slm.finishInitialization();
TaskBatch batch = new TaskBatch();
String tasknode = submitTaskAndWait(batch, "foo/1");
ZKUtil.setData(zkw, tasknode, TaskState.TASK_ERR.get("worker"));
synchronized (batch) {
while (batch.installed != batch.error) {
batch.wait();
}
}
waitForCounter(tot_mgr_task_deleted, 0, 1, 1000);
assertTrue(ZKUtil.checkExists(zkw, tasknode) == -1);
conf.setInt("hbase.splitlog.max.resubmit", ZKSplitLog.DEFAULT_MAX_RESUBMIT);
}
@Test
public void testUnassignedTimeout() throws Exception {
LOG.info("TestUnassignedTimeout - iff all tasks are unassigned then" + " resubmit");
// create an orphan task in OWNED state
String tasknode1 = ZKSplitLog.getEncodedNodeName(zkw, "orphan/1");
zkw.getRecoverableZooKeeper()
.create(
tasknode1,
TaskState.TASK_OWNED.get("dummy-worker"),
Ids.OPEN_ACL_UNSAFE,
CreateMode.PERSISTENT);
int to = 1000;
conf.setInt("hbase.splitlog.manager.timeout", to);
conf.setInt("hbase.splitlog.manager.unassigned.timeout", 2 * to);
conf.setInt("hbase.splitlog.manager.timeoutmonitor.period", 100);
slm = new SplitLogManager(zkw, conf, stopper, "dummy-master", null);
slm.finishInitialization();
waitForCounter(tot_mgr_orphan_task_acquired, 0, 1, 100);
// submit another task which will stay in unassigned mode
TaskBatch batch = new TaskBatch();
submitTaskAndWait(batch, "foo/1");
// keep updating the orphan owned node every to/2 seconds
for (int i = 0; i < (3 * to) / 100; i++) {
Thread.sleep(100);
ZKUtil.setData(zkw, tasknode1, TaskState.TASK_OWNED.get("dummy-worker"));
}
// since we have stopped heartbeating the owned node therefore it should
// get resubmitted
LOG.info("waiting for manager to resubmit the orphan task");
waitForCounter(tot_mgr_resubmit, 0, 1, to + 500);
// now all the nodes are unassigned. manager should post another rescan
waitForCounter(tot_mgr_resubmit_unassigned, 0, 1, 2 * to + 500);
}
/**
* Update the state znode of a peer cluster.
*
* @param id
* @param state
*/
private void changePeerState(String id, ZooKeeperProtos.ReplicationState.State state)
throws ReplicationException {
try {
if (!peerExists(id)) {
throw new IllegalArgumentException(
"Cannot enable/disable peer because id=" + id + " does not exist.");
}
String peerStateZNode = getPeerStateNode(id);
byte[] stateBytes =
(state == ZooKeeperProtos.ReplicationState.State.ENABLED)
? ENABLED_ZNODE_BYTES
: DISABLED_ZNODE_BYTES;
if (ZKUtil.checkExists(this.zookeeper, peerStateZNode) != -1) {
ZKUtil.setData(this.zookeeper, peerStateZNode, stateBytes);
} else {
ZKUtil.createAndWatch(this.zookeeper, peerStateZNode, stateBytes);
}
LOG.info("Peer with id= " + id + " is now " + state.name());
} catch (KeeperException e) {
throw new ReplicationException("Unable to change state of the peer with id=" + id, e);
}
}
@Test
public void testRescanCleanup() throws Exception {
LOG.info("TestRescanCleanup - ensure RESCAN nodes are cleaned up");
conf.setInt("hbase.splitlog.manager.timeout", 1000);
conf.setInt("hbase.splitlog.manager.timeoutmonitor.period", 100);
slm = new SplitLogManager(zkw, conf, stopper, "dummy-master", null);
slm.finishInitialization();
TaskBatch batch = new TaskBatch();
String tasknode = submitTaskAndWait(batch, "foo/1");
int version = ZKUtil.checkExists(zkw, tasknode);
ZKUtil.setData(zkw, tasknode, TaskState.TASK_OWNED.get("worker1"));
waitForCounter(tot_mgr_heartbeat, 0, 1, 1000);
waitForCounter(
new Expr() {
@Override
public long eval() {
return (tot_mgr_resubmit.get() + tot_mgr_resubmit_failed.get());
}
},
0,
1,
5 * 60000); // wait long enough
if (tot_mgr_resubmit_failed.get() == 0) {
int version1 = ZKUtil.checkExists(zkw, tasknode);
assertTrue(version1 > version);
byte[] taskstate = ZKUtil.getData(zkw, tasknode);
assertTrue(Arrays.equals(TaskState.TASK_UNASSIGNED.get("dummy-master"), taskstate));
waitForCounter(tot_mgr_rescan_deleted, 0, 1, 1000);
} else {
LOG.warn("Could not run test. Lost ZK connection?");
}
return;
}
public boolean shouldAddCheckerMaster() {
ZooKeeperWatcher zk = super.getZooKeeper();
String numberN = ZKUtil.joinZNode(zk.baseZNode, CCIndexConstants.CheckNumNode);
try {
if (ZKUtil.checkExists(zk, numberN) != -1) {
ZKUtil.createSetData(zk, numberN, Bytes.toBytes(1));
} else {
int num = Bytes.toInt(ZKUtil.getData(zk, numberN));
if (num < this.checkMasterN) {
ZKUtil.setData(zk, numberN, Bytes.toBytes(num + 1));
return true;
} else {
return false;
}
}
} catch (KeeperException e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
return false;
}
@Test
public void testTaskErr() throws Exception {
LOG.info("TestTaskErr - cleanup task node once in ERR state");
conf.setInt("hbase.splitlog.max.resubmit", 0);
slm = new SplitLogManager(zkw, conf, stopper, master, DUMMY_MASTER, null);
slm.finishInitialization();
TaskBatch batch = new TaskBatch();
String tasknode = submitTaskAndWait(batch, "foo/1");
final ServerName worker1 = new ServerName("worker1,1,1");
SplitLogTask slt = new SplitLogTask.Err(worker1);
ZKUtil.setData(zkw, tasknode, slt.toByteArray());
synchronized (batch) {
while (batch.installed != batch.error) {
batch.wait();
}
}
waitForCounter(tot_mgr_task_deleted, 0, 1, to / 2);
assertTrue(ZKUtil.checkExists(zkw, tasknode) == -1);
conf.setInt("hbase.splitlog.max.resubmit", SplitLogManager.DEFAULT_MAX_RESUBMIT);
}
@Test
public void testDeadWorker() throws Exception {
LOG.info("testDeadWorker");
conf.setLong("hbase.splitlog.max.resubmit", 0);
slm = new SplitLogManager(zkw, conf, stopper, "dummy-master", null);
slm.finishInitialization();
TaskBatch batch = new TaskBatch();
String tasknode = submitTaskAndWait(batch, "foo/1");
int version = ZKUtil.checkExists(zkw, tasknode);
ZKUtil.setData(zkw, tasknode, TaskState.TASK_OWNED.get("worker1"));
waitForCounter(tot_mgr_heartbeat, 0, 1, 1000);
slm.handleDeadWorker("worker1");
waitForCounter(tot_mgr_resubmit, 0, 1, 1000);
waitForCounter(tot_mgr_resubmit_dead_server_task, 0, 1, 1000);
int version1 = ZKUtil.checkExists(zkw, tasknode);
assertTrue(version1 > version);
byte[] taskstate = ZKUtil.getData(zkw, tasknode);
assertTrue(Arrays.equals(TaskState.TASK_UNASSIGNED.get("dummy-master"), taskstate));
return;
}
@Test
public void testWorkerCrash() throws Exception {
slm = new SplitLogManager(zkw, conf, stopper, master, DUMMY_MASTER, null);
slm.finishInitialization();
TaskBatch batch = new TaskBatch();
String tasknode = submitTaskAndWait(batch, "foo/1");
final ServerName worker1 = new ServerName("worker1,1,1");
SplitLogTask slt = new SplitLogTask.Owned(worker1);
ZKUtil.setData(zkw, tasknode, slt.toByteArray());
if (tot_mgr_heartbeat.get() == 0) waitForCounter(tot_mgr_heartbeat, 0, 1, to / 2);
// Not yet resubmitted.
Assert.assertEquals(0, tot_mgr_resubmit.get());
// This server becomes dead
Mockito.when(sm.isServerOnline(worker1)).thenReturn(false);
Thread.sleep(1300); // The timeout checker is done every 1000 ms (hardcoded).
// It has been resubmitted
Assert.assertEquals(1, tot_mgr_resubmit.get());
}
