use of org.apache.kafka.connect.util.ConnectorTaskId in project kafka by apache.
the class IncrementalCooperativeAssignor method performTaskRevocation.
/**
* Task revocation is based on an rough estimation of the lower average number of tasks before
* and after new workers join the group. If no new workers join, no revocation takes place.
* Based on this estimation, tasks are revoked until the new floor average is reached for
* each existing worker. The revoked tasks, once assigned to the new workers will maintain
* a balanced load among the group.
*
* @param activeAssignments
* @param completeWorkerAssignment
* @return
*/
private Map<String, ConnectorsAndTasks> performTaskRevocation(ConnectorsAndTasks activeAssignments, Collection<WorkerLoad> completeWorkerAssignment) {
int totalActiveConnectorsNum = activeAssignments.connectors().size();
int totalActiveTasksNum = activeAssignments.tasks().size();
Collection<WorkerLoad> existingWorkers = completeWorkerAssignment.stream().filter(wl -> wl.size() > 0).collect(Collectors.toList());
int existingWorkersNum = existingWorkers.size();
int totalWorkersNum = completeWorkerAssignment.size();
int newWorkersNum = totalWorkersNum - existingWorkersNum;
if (log.isDebugEnabled()) {
completeWorkerAssignment.forEach(wl -> log.debug("Per worker current load size; worker: {} connectors: {} tasks: {}", wl.worker(), wl.connectorsSize(), wl.tasksSize()));
}
Map<String, ConnectorsAndTasks> revoking = new HashMap<>();
// after logging the status
if (!(newWorkersNum > 0 && existingWorkersNum > 0)) {
log.debug("No task revocation required; workers with existing load: {} workers with " + "no load {} total workers {}", existingWorkersNum, newWorkersNum, totalWorkersNum);
// connectors and tasks as well
return revoking;
}
log.debug("Task revocation is required; workers with existing load: {} workers with " + "no load {} total workers {}", existingWorkersNum, newWorkersNum, totalWorkersNum);
// We have at least one worker assignment (the leader itself) so totalWorkersNum can't be 0
log.debug("Previous rounded down (floor) average number of connectors per worker {}", totalActiveConnectorsNum / existingWorkersNum);
int floorConnectors = totalActiveConnectorsNum / totalWorkersNum;
int ceilConnectors = floorConnectors + ((totalActiveConnectorsNum % totalWorkersNum == 0) ? 0 : 1);
log.debug("New average number of connectors per worker rounded down (floor) {} and rounded up (ceil) {}", floorConnectors, ceilConnectors);
log.debug("Previous rounded down (floor) average number of tasks per worker {}", totalActiveTasksNum / existingWorkersNum);
int floorTasks = totalActiveTasksNum / totalWorkersNum;
int ceilTasks = floorTasks + ((totalActiveTasksNum % totalWorkersNum == 0) ? 0 : 1);
log.debug("New average number of tasks per worker rounded down (floor) {} and rounded up (ceil) {}", floorTasks, ceilTasks);
int numToRevoke;
for (WorkerLoad existing : existingWorkers) {
Iterator<String> connectors = existing.connectors().iterator();
numToRevoke = existing.connectorsSize() - ceilConnectors;
for (int i = existing.connectorsSize(); i > floorConnectors && numToRevoke > 0; --i, --numToRevoke) {
ConnectorsAndTasks resources = revoking.computeIfAbsent(existing.worker(), w -> new ConnectorsAndTasks.Builder().build());
resources.connectors().add(connectors.next());
}
}
for (WorkerLoad existing : existingWorkers) {
Iterator<ConnectorTaskId> tasks = existing.tasks().iterator();
numToRevoke = existing.tasksSize() - ceilTasks;
log.debug("Tasks on worker {} is higher than ceiling, so revoking {} tasks", existing, numToRevoke);
for (int i = existing.tasksSize(); i > floorTasks && numToRevoke > 0; --i, --numToRevoke) {
ConnectorsAndTasks resources = revoking.computeIfAbsent(existing.worker(), w -> new ConnectorsAndTasks.Builder().build());
resources.tasks().add(tasks.next());
}
}
return revoking;
}
use of org.apache.kafka.connect.util.ConnectorTaskId in project kafka by apache.
the class IncrementalCooperativeAssignor method assignTasks.
/**
* Perform a round-robin assignment of tasks to workers with existing worker load. This
* assignment tries to balance the load between workers, by assigning tasks to workers that
* have equal load, starting with the least loaded workers.
*
* @param workerAssignment the current worker assignment; assigned tasks are added to this list
* @param tasks the tasks to be assigned
*/
protected void assignTasks(List<WorkerLoad> workerAssignment, Collection<ConnectorTaskId> tasks) {
workerAssignment.sort(WorkerLoad.taskComparator());
WorkerLoad first = workerAssignment.get(0);
Iterator<ConnectorTaskId> load = tasks.iterator();
while (load.hasNext()) {
int firstLoad = first.tasksSize();
int upTo = IntStream.range(0, workerAssignment.size()).filter(i -> workerAssignment.get(i).tasksSize() > firstLoad).findFirst().orElse(workerAssignment.size());
for (WorkerLoad worker : workerAssignment.subList(0, upTo)) {
ConnectorTaskId task = load.next();
log.debug("Assigning task {} to {}", task, worker.worker());
worker.assign(task);
if (!load.hasNext()) {
break;
}
}
}
}
use of org.apache.kafka.connect.util.ConnectorTaskId in project kafka by apache.
the class EagerAssignor method fillAssignmentsAndSerialize.
private Map<String, ByteBuffer> fillAssignmentsAndSerialize(Collection<String> members, short error, String leaderId, String leaderUrl, long maxOffset, Map<String, Collection<String>> connectorAssignments, Map<String, Collection<ConnectorTaskId>> taskAssignments) {
Map<String, ByteBuffer> groupAssignment = new HashMap<>();
for (String member : members) {
Collection<String> connectors = connectorAssignments.get(member);
if (connectors == null) {
connectors = Collections.emptyList();
}
Collection<ConnectorTaskId> tasks = taskAssignments.get(member);
if (tasks == null) {
tasks = Collections.emptyList();
}
Assignment assignment = new Assignment(error, leaderId, leaderUrl, maxOffset, connectors, tasks);
log.debug("Assignment: {} -> {}", member, assignment);
groupAssignment.put(member, ConnectProtocol.serializeAssignment(assignment));
}
log.debug("Finished assignment");
return groupAssignment;
}
use of org.apache.kafka.connect.util.ConnectorTaskId in project kafka by apache.
the class IncrementalCooperativeAssignor method handleLostAssignments.
// visible for testing
protected void handleLostAssignments(ConnectorsAndTasks lostAssignments, ConnectorsAndTasks newSubmissions, List<WorkerLoad> completeWorkerAssignment, Map<String, ExtendedWorkerState> memberConfigs) {
if (lostAssignments.isEmpty()) {
resetDelay();
return;
}
final long now = time.milliseconds();
log.debug("Found the following connectors and tasks missing from previous assignments: " + lostAssignments);
if (scheduledRebalance <= 0 && memberConfigs.keySet().containsAll(previousMembers)) {
log.debug("No worker seems to have departed the group during the rebalance. The " + "missing assignments that the leader is detecting are probably due to some " + "workers failing to receive the new assignments in the previous rebalance. " + "Will reassign missing tasks as new tasks");
newSubmissions.connectors().addAll(lostAssignments.connectors());
newSubmissions.tasks().addAll(lostAssignments.tasks());
return;
}
if (scheduledRebalance > 0 && now >= scheduledRebalance) {
// delayed rebalance expired and it's time to assign resources
log.debug("Delayed rebalance expired. Reassigning lost tasks");
List<WorkerLoad> candidateWorkerLoad = Collections.emptyList();
if (!candidateWorkersForReassignment.isEmpty()) {
candidateWorkerLoad = pickCandidateWorkerForReassignment(completeWorkerAssignment);
}
if (!candidateWorkerLoad.isEmpty()) {
log.debug("Assigning lost tasks to {} candidate workers: {}", candidateWorkerLoad.size(), candidateWorkerLoad.stream().map(WorkerLoad::worker).collect(Collectors.joining(",")));
Iterator<WorkerLoad> candidateWorkerIterator = candidateWorkerLoad.iterator();
for (String connector : lostAssignments.connectors()) {
// Loop over the candidate workers as many times as it takes
if (!candidateWorkerIterator.hasNext()) {
candidateWorkerIterator = candidateWorkerLoad.iterator();
}
WorkerLoad worker = candidateWorkerIterator.next();
log.debug("Assigning connector id {} to member {}", connector, worker.worker());
worker.assign(connector);
}
candidateWorkerIterator = candidateWorkerLoad.iterator();
for (ConnectorTaskId task : lostAssignments.tasks()) {
if (!candidateWorkerIterator.hasNext()) {
candidateWorkerIterator = candidateWorkerLoad.iterator();
}
WorkerLoad worker = candidateWorkerIterator.next();
log.debug("Assigning task id {} to member {}", task, worker.worker());
worker.assign(task);
}
} else {
log.debug("No single candidate worker was found to assign lost tasks. Treating lost tasks as new tasks");
newSubmissions.connectors().addAll(lostAssignments.connectors());
newSubmissions.tasks().addAll(lostAssignments.tasks());
}
resetDelay();
} else {
candidateWorkersForReassignment.addAll(candidateWorkersForReassignment(completeWorkerAssignment));
if (now < scheduledRebalance) {
// a delayed rebalance is in progress, but it's not yet time to reassign
// unaccounted resources
delay = calculateDelay(now);
log.debug("Delayed rebalance in progress. Task reassignment is postponed. New computed rebalance delay: {}", delay);
} else {
// This means scheduledRebalance == 0
// We could also also extract the current minimum delay from the group, to make
// independent of consecutive leader failures, but this optimization is skipped
// at the moment
delay = maxDelay;
log.debug("Resetting rebalance delay to the max: {}. scheduledRebalance: {} now: {} diff scheduledRebalance - now: {}", delay, scheduledRebalance, now, scheduledRebalance - now);
}
scheduledRebalance = now + delay;
}
}
use of org.apache.kafka.connect.util.ConnectorTaskId in project kafka by apache.
the class AbstractHerderTest method testConnectorStatus.
@Test
public void testConnectorStatus() {
ConnectorTaskId taskId = new ConnectorTaskId(connector, 0);
AbstractHerder herder = partialMockBuilder(AbstractHerder.class).withConstructor(Worker.class, String.class, String.class, StatusBackingStore.class, ConfigBackingStore.class, ConnectorClientConfigOverridePolicy.class).withArgs(worker, workerId, kafkaClusterId, statusStore, configStore, noneConnectorClientConfigOverridePolicy).addMockedMethod("generation").createMock();
EasyMock.expect(herder.generation()).andStubReturn(generation);
EasyMock.expect(herder.rawConfig(connector)).andReturn(null);
EasyMock.expect(statusStore.get(connector)).andReturn(new ConnectorStatus(connector, AbstractStatus.State.RUNNING, workerId, generation));
EasyMock.expect(statusStore.getAll(connector)).andReturn(Collections.singletonList(new TaskStatus(taskId, AbstractStatus.State.UNASSIGNED, workerId, generation)));
replayAll();
ConnectorStateInfo csi = herder.connectorStatus(connector);
PowerMock.verifyAll();
}
Aggregations