Examples with TopologyChangedException com.hazelcast.jet.core.TopologyChangedException used on opensource projects

Example 1 with TopologyChangedException

use of com.hazelcast.jet.core.TopologyChangedException in project hazelcast-jet by hazelcast.

the class MasterContext method getExecuteResult.

/**
 * <ul>
 * <li>Returns null if there is no failure.
 * <li>Returns CancellationException if the job is cancelled.
 * <li>Returns JobRestartRequestedException if the current execution is cancelled
 * <li>If there is at least one non-restartable failure, such as an exception in user code, then returns that failure.
 * <li>Otherwise, the failure is because a job participant has left the cluster.
 *   In that case, {@code TopologyChangeException} is returned so that the job will be restarted.
 * </ul>
 */
private Throwable getExecuteResult(Map<MemberInfo, Object> responses, boolean isRestartRequested) {
    if (cancellationToken.isCompleted()) {
        logger.fine(jobIdString() + " to be cancelled after execute");
        return new CancellationException();
    } else if (isRestartRequested) {
        return new JobRestartRequestedException();
    }
    Map<Boolean, List<Entry<MemberInfo, Object>>> grouped = groupResponses(responses);
    Collection<MemberInfo> successfulMembers = grouped.get(false).stream().map(Entry::getKey).collect(toList());
    if (successfulMembers.size() == executionPlanMap.size()) {
        logger.fine("Execute of " + jobIdString() + " is successful.");
        return null;
    }
    List<Entry<MemberInfo, Object>> failures = grouped.get(true);
    logger.fine("Execute of " + jobIdString() + " has failures: " + failures);
    // In that case, all remaining participants return a CancellationException.
    return failures.stream().map(e -> (Throwable) e.getValue()).filter(t -> !(t instanceof CancellationException || isTopologicalFailure(t))).findFirst().map(ExceptionUtil::peel).orElse(new TopologyChangedException());
}

Example 2 with TopologyChangedException

use of com.hazelcast.jet.core.TopologyChangedException in project hazelcast-jet by hazelcast.

the class JobCoordinationService method completeMasterContextIfJobAlreadyCompleted.

// If a job result is present, it completes the master context using the job result
private boolean completeMasterContextIfJobAlreadyCompleted(MasterContext masterContext) {
    long jobId = masterContext.getJobId();
    JobResult jobResult = jobRepository.getJobResult(jobId);
    if (jobResult != null) {
        logger.fine("Completing master context " + idToString(jobId) + " since already completed with result: " + jobResult);
        masterContext.setFinalResult(jobResult.getFailure());
        return masterContexts.remove(jobId, masterContext);
    }
    if (!masterContext.getJobConfig().isAutoRestartOnMemberFailureEnabled() && jobRepository.getExecutionIdCount(jobId) > 0) {
        String coordinator = nodeEngine.getNode().getThisUuid();
        Throwable result = new TopologyChangedException();
        logger.info("Completing Job " + idToString(jobId) + " with " + result + " since auto-restart is disabled and the job has been executed before");
        jobRepository.completeJob(jobId, coordinator, System.currentTimeMillis(), result);
        masterContext.setFinalResult(result);
        return masterContexts.remove(jobId, masterContext);
    }
    return false;
}

Example 3 with TopologyChangedException

use of com.hazelcast.jet.core.TopologyChangedException in project hazelcast-jet by hazelcast.

the class JobExecutionService method verifyClusterInformation.

private void verifyClusterInformation(long jobId, long executionId, Address coordinator, int coordinatorMemberListVersion, Set<MemberInfo> participants) {
    Address masterAddress = nodeEngine.getMasterAddress();
    if (!coordinator.equals(masterAddress)) {
        failIfNotRunning();
        throw new IllegalStateException(String.format("Coordinator %s cannot initialize %s. Reason: it is not the master, the master is %s", coordinator, jobAndExecutionId(jobId, executionId), masterAddress));
    }
    ClusterServiceImpl clusterService = (ClusterServiceImpl) nodeEngine.getClusterService();
    MembershipManager membershipManager = clusterService.getMembershipManager();
    int localMemberListVersion = membershipManager.getMemberListVersion();
    Address thisAddress = nodeEngine.getThisAddress();
    if (coordinatorMemberListVersion > localMemberListVersion) {
        assert !masterAddress.equals(thisAddress) : String.format("Local node: %s is master but InitOperation has coordinator member list version: %s larger than " + " local member list version: %s", thisAddress, coordinatorMemberListVersion, localMemberListVersion);
        nodeEngine.getOperationService().send(new TriggerMemberListPublishOp(), masterAddress);
        throw new RetryableHazelcastException(String.format("Cannot initialize %s for coordinator %s, local member list version %s," + " coordinator member list version %s", jobAndExecutionId(jobId, executionId), coordinator, localMemberListVersion, coordinatorMemberListVersion));
    }
    boolean isLocalMemberParticipant = false;
    for (MemberInfo participant : participants) {
        if (participant.getAddress().equals(thisAddress)) {
            isLocalMemberParticipant = true;
        }
        if (membershipManager.getMember(participant.getAddress(), participant.getUuid()) == null) {
            throw new TopologyChangedException(String.format("Cannot initialize %s for coordinator %s: participant %s not found in local member list." + " Local member list version: %s, coordinator member list version: %s", jobAndExecutionId(jobId, executionId), coordinator, participant, localMemberListVersion, coordinatorMemberListVersion));
        }
    }
    if (!isLocalMemberParticipant) {
        throw new IllegalArgumentException(String.format("Cannot initialize %s since member %s is not in participants: %s", jobAndExecutionId(jobId, executionId), thisAddress, participants));
    }
}

Example 4 with TopologyChangedException

use of com.hazelcast.jet.core.TopologyChangedException in project hazelcast-jet by hazelcast.

the class JobExecutionService method assertExecutionContext.

public ExecutionContext assertExecutionContext(Address coordinator, long jobId, long executionId, Operation callerOp) {
    Address masterAddress = nodeEngine.getMasterAddress();
    if (!coordinator.equals(masterAddress)) {
        failIfNotRunning();
        throw new IllegalStateException(String.format("Coordinator %s cannot do '%s' for %s: it is not the master, the master is %s", coordinator, callerOp.getClass().getSimpleName(), jobAndExecutionId(jobId, executionId), masterAddress));
    }
    failIfNotRunning();
    ExecutionContext executionContext = executionContexts.get(executionId);
    if (executionContext == null) {
        throw new TopologyChangedException(String.format("%s not found for coordinator %s for '%s'", jobAndExecutionId(jobId, executionId), coordinator, callerOp.getClass().getSimpleName()));
    } else if (!(executionContext.coordinator().equals(coordinator) && executionContext.jobId() == jobId)) {
        throw new IllegalStateException(String.format("%s, originally from coordinator %s, cannot do '%s' by coordinator %s and execution %s", jobAndExecutionId(jobId, executionContext.executionId()), executionContext.coordinator(), callerOp.getClass().getSimpleName(), coordinator, idToString(executionId)));
    }
    return executionContext;
}

Example 5 with TopologyChangedException

use of com.hazelcast.jet.core.TopologyChangedException in project hazelcast by hazelcast.

the class JobCoordinationService method completeMasterContextIfJobAlreadyCompleted.

// If a job result is present, it completes the master context using the job result
private boolean completeMasterContextIfJobAlreadyCompleted(MasterContext masterContext) {
    long jobId = masterContext.jobId();
    JobResult jobResult = jobRepository.getJobResult(jobId);
    if (jobResult != null) {
        logger.fine("Completing master context for " + masterContext.jobIdString() + " since already completed with result: " + jobResult);
        masterContext.jobContext().setFinalResult(jobResult.getFailureAsThrowable());
        return masterContexts.remove(jobId, masterContext);
    }
    if (!masterContext.jobConfig().isAutoScaling() && masterContext.jobExecutionRecord().executed()) {
        logger.info("Suspending or failing " + masterContext.jobIdString() + " since auto-restart is disabled and the job has been executed before");
        masterContext.jobContext().finalizeJob(new TopologyChangedException());
        return true;
    }
    return false;
}