@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());
}
@Test
public void testOrphanTaskAcquisition() throws Exception {
LOG.info("TestOrphanTaskAcquisition");
String tasknode = ZKSplitLog.getEncodedNodeName(zkw, "orphan/test/slash");
zkw.getRecoverableZooKeeper()
.create(
tasknode,
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.timeoutmonitor.period", 100);
to = to + 2 * 100;
slm = new SplitLogManager(zkw, conf, stopper, "dummy-master", null);
slm.finishInitialization();
waitForCounter(tot_mgr_orphan_task_acquired, 0, 1, 100);
Task task = slm.findOrCreateOrphanTask(tasknode);
assertTrue(task.isOrphan());
waitForCounter(tot_mgr_heartbeat, 0, 1, 100);
assertFalse(task.isUnassigned());
long curt = System.currentTimeMillis();
assertTrue((task.last_update <= curt) && (task.last_update > (curt - 1000)));
LOG.info("waiting for manager to resubmit the orphan task");
waitForCounter(tot_mgr_resubmit, 0, 1, to + 100);
assertTrue(task.isUnassigned());
waitForCounter(tot_mgr_rescan, 0, 1, to + 100);
}
@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);
}
@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;
}
@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;
}