Examples with ErrorCode io.trino.spi.ErrorCode used on opensource projects

Example 1 with ErrorCode

use of io.trino.spi.ErrorCode in project trino by trinodb.

the class QueryStateMachine method getBasicQueryInfo.

public BasicQueryInfo getBasicQueryInfo(Optional<BasicStageStats> rootStage) {
    // Query state must be captured first in order to provide a
    // correct view of the query.  For example, building this
    // information, the query could finish, and the task states would
    // never be visible.
    QueryState state = queryState.get();
    ErrorCode errorCode = null;
    if (state == FAILED) {
        ExecutionFailureInfo failureCause = this.failureCause.get();
        if (failureCause != null) {
            errorCode = failureCause.getErrorCode();
        }
    }
    BasicStageStats stageStats = rootStage.orElse(EMPTY_STAGE_STATS);
    BasicQueryStats queryStats = new BasicQueryStats(queryStateTimer.getCreateTime(), getEndTime().orElse(null), queryStateTimer.getQueuedTime(), queryStateTimer.getElapsedTime(), queryStateTimer.getExecutionTime(), stageStats.getTotalDrivers(), stageStats.getQueuedDrivers(), stageStats.getRunningDrivers(), stageStats.getCompletedDrivers(), stageStats.getRawInputDataSize(), stageStats.getRawInputPositions(), stageStats.getPhysicalInputDataSize(), stageStats.getCumulativeUserMemory(), stageStats.getFailedCumulativeUserMemory(), stageStats.getUserMemoryReservation(), stageStats.getTotalMemoryReservation(), succinctBytes(getPeakUserMemoryInBytes()), succinctBytes(getPeakTotalMemoryInBytes()), stageStats.getTotalCpuTime(), stageStats.getFailedCpuTime(), stageStats.getTotalScheduledTime(), stageStats.getFailedScheduledTime(), stageStats.isFullyBlocked(), stageStats.getBlockedReasons(), stageStats.getProgressPercentage());
    return new BasicQueryInfo(queryId, session.toSessionRepresentation(), Optional.of(resourceGroup), state, stageStats.isScheduled(), self, query, Optional.ofNullable(updateType.get()), preparedQuery, queryStats, errorCode == null ? null : errorCode.getType(), errorCode, queryType);
}

Example 2 with ErrorCode

use of io.trino.spi.ErrorCode in project trino by trinodb.

the class QueryStateMachine method getQueryInfo.

@VisibleForTesting
QueryInfo getQueryInfo(Optional<StageInfo> rootStage) {
    // Query state must be captured first in order to provide a
    // correct view of the query.  For example, building this
    // information, the query could finish, and the task states would
    // never be visible.
    QueryState state = queryState.get();
    ExecutionFailureInfo failureCause = null;
    ErrorCode errorCode = null;
    if (state == FAILED) {
        failureCause = this.failureCause.get();
        if (failureCause != null) {
            errorCode = failureCause.getErrorCode();
        }
    }
    boolean completeInfo = getAllStages(rootStage).stream().allMatch(StageInfo::isCompleteInfo);
    boolean isScheduled = isScheduled(rootStage);
    return new QueryInfo(queryId, session.toSessionRepresentation(), state, isScheduled, self, outputManager.getQueryOutputInfo().map(QueryOutputInfo::getColumnNames).orElse(ImmutableList.of()), query, preparedQuery, getQueryStats(rootStage), Optional.ofNullable(setCatalog.get()), Optional.ofNullable(setSchema.get()), Optional.ofNullable(setPath.get()), setSessionProperties, resetSessionProperties, setRoles, addedPreparedStatements, deallocatedPreparedStatements, Optional.ofNullable(startedTransactionId.get()), clearTransactionId.get(), updateType.get(), rootStage, failureCause, errorCode, warningCollector.getWarnings(), inputs.get(), output.get(), referencedTables.get(), routines.get(), completeInfo, Optional.of(resourceGroup), queryType);
}

Example 3 with ErrorCode

use of io.trino.spi.ErrorCode in project trino by trinodb.

the class FaultTolerantStageScheduler method updateTaskStatus.

private void updateTaskStatus(TaskStatus taskStatus, Optional<ExchangeSinkInstanceHandle> exchangeSinkInstanceHandle) {
    TaskState state = taskStatus.getState();
    if (!state.isDone()) {
        return;
    }
    try {
        RuntimeException failure = null;
        SettableFuture<Void> future;
        synchronized (this) {
            TaskId taskId = taskStatus.getTaskId();
            runningTasks.remove(taskId);
            future = taskFinishedFuture;
            if (!runningTasks.isEmpty()) {
                taskFinishedFuture = SettableFuture.create();
            } else {
                taskFinishedFuture = null;
            }
            NodeAllocator.NodeLease nodeLease = requireNonNull(runningNodes.remove(taskId), () -> "node not found for task id: " + taskId);
            nodeLease.release();
            int partitionId = taskId.getPartitionId();
            if (!finishedPartitions.contains(partitionId) && !closed) {
                switch(state) {
                    case FINISHED:
                        finishedPartitions.add(partitionId);
                        if (sinkExchange.isPresent()) {
                            checkArgument(exchangeSinkInstanceHandle.isPresent(), "exchangeSinkInstanceHandle is expected to be present");
                            sinkExchange.get().sinkFinished(exchangeSinkInstanceHandle.get());
                        }
                        partitionToRemoteTaskMap.get(partitionId).forEach(RemoteTask::abort);
                        break;
                    case CANCELED:
                        log.debug("Task cancelled: %s", taskId);
                        break;
                    case ABORTED:
                        log.debug("Task aborted: %s", taskId);
                        break;
                    case FAILED:
                        ExecutionFailureInfo failureInfo = taskStatus.getFailures().stream().findFirst().map(this::rewriteTransportFailure).orElse(toFailure(new TrinoException(GENERIC_INTERNAL_ERROR, "A task failed for an unknown reason")));
                        log.warn(failureInfo.toException(), "Task failed: %s", taskId);
                        ErrorCode errorCode = failureInfo.getErrorCode();
                        int taskRemainingAttempts = remainingAttemptsPerTask.getOrDefault(partitionId, maxRetryAttemptsPerTask);
                        if (remainingRetryAttemptsOverall > 0 && taskRemainingAttempts > 0 && (errorCode == null || errorCode.getType() != USER_ERROR)) {
                            remainingRetryAttemptsOverall--;
                            remainingAttemptsPerTask.put(partitionId, taskRemainingAttempts - 1);
                            // update memory limits for next attempt
                            MemoryRequirements memoryLimits = partitionMemoryRequirements.get(partitionId);
                            verify(memoryLimits != null);
                            MemoryRequirements newMemoryLimits = partitionMemoryEstimator.getNextRetryMemoryRequirements(session, memoryLimits, errorCode);
                            partitionMemoryRequirements.put(partitionId, newMemoryLimits);
                            // reschedule
                            queuedPartitions.add(partitionId);
                            log.debug("Retrying partition %s for stage %s", partitionId, stage.getStageId());
                        } else {
                            failure = failureInfo.toException();
                        }
                        break;
                    default:
                        throw new IllegalArgumentException("Unexpected task state: " + state);
                }
            }
        }
        if (failure != null) {
            // must be called outside the lock
            fail(failure);
        }
        if (future != null && !future.isDone()) {
            future.set(null);
        }
    } catch (Throwable t) {
        fail(t);
    }
}

Example 4 with ErrorCode

use of io.trino.spi.ErrorCode in project trino by trinodb.

the class SqlQueryScheduler method createDistributedStagesScheduler.

private synchronized Optional<DistributedStagesScheduler> createDistributedStagesScheduler(int attempt) {
    verify(attempt == 0 || retryPolicy == RetryPolicy.QUERY, "unexpected attempt %s for retry policy %s", attempt, retryPolicy);
    if (queryStateMachine.isDone()) {
        return Optional.empty();
    }
    if (attempt > 0 && retryPolicy == RetryPolicy.QUERY) {
        dynamicFilterService.registerQueryRetry(queryStateMachine.getQueryId(), attempt);
    }
    DistributedStagesScheduler distributedStagesScheduler;
    switch(retryPolicy) {
        case TASK:
            ExchangeManager exchangeManager = exchangeManagerRegistry.getExchangeManager();
            distributedStagesScheduler = FaultTolerantDistributedStagesScheduler.create(queryStateMachine, stageManager, failureDetector, taskSourceFactory, taskDescriptorStorage, exchangeManager, nodePartitioningManager, coordinatorStagesScheduler.getTaskLifecycleListener(), maxTaskRetryAttemptsOverall, maxTaskRetryAttemptsPerTask, schedulerExecutor, schedulerStats, nodeAllocatorService, partitionMemoryEstimator);
            break;
        case QUERY:
        case NONE:
            distributedStagesScheduler = PipelinedDistributedStagesScheduler.create(queryStateMachine, schedulerStats, nodeScheduler, nodePartitioningManager, stageManager, coordinatorStagesScheduler, executionPolicy, failureDetector, schedulerExecutor, splitSourceFactory, splitBatchSize, dynamicFilterService, tableExecuteContextManager, retryPolicy, attempt);
            break;
        default:
            throw new IllegalArgumentException("Unexpected retry policy: " + retryPolicy);
    }
    this.distributedStagesScheduler.set(distributedStagesScheduler);
    distributedStagesScheduler.addStateChangeListener(state -> {
        if (queryStateMachine.getQueryState() == QueryState.STARTING && (state == DistributedStagesSchedulerState.RUNNING || state.isDone())) {
            queryStateMachine.transitionToRunning();
        }
        if (state.isDone() && !state.isFailure()) {
            stageManager.getDistributedStagesInTopologicalOrder().forEach(stage -> stageManager.get(stage.getStageId()).finish());
        }
        if (stageManager.getCoordinatorStagesInTopologicalOrder().isEmpty()) {
            // otherwise defer query transitioning to the coordinator stages
            if (state == DistributedStagesSchedulerState.FINISHED) {
                queryStateMachine.transitionToFinishing();
            } else if (state == DistributedStagesSchedulerState.CANCELED) {
                // output stage was canceled
                queryStateMachine.transitionToCanceled();
            }
        }
        if (state == DistributedStagesSchedulerState.FAILED) {
            StageFailureInfo stageFailureInfo = distributedStagesScheduler.getFailureCause().orElseGet(() -> new StageFailureInfo(toFailure(new VerifyException("distributedStagesScheduler failed but failure cause is not present")), Optional.empty()));
            ErrorCode errorCode = stageFailureInfo.getFailureInfo().getErrorCode();
            if (shouldRetry(errorCode)) {
                long delayInMillis = min(retryInitialDelay.toMillis() * ((long) pow(2, currentAttempt.get())), retryMaxDelay.toMillis());
                currentAttempt.incrementAndGet();
                scheduleRetryWithDelay(delayInMillis);
            } else {
                stageManager.getDistributedStagesInTopologicalOrder().forEach(stage -> {
                    if (stageFailureInfo.getFailedStageId().isPresent() && stageFailureInfo.getFailedStageId().get().equals(stage.getStageId())) {
                        stage.fail(stageFailureInfo.getFailureInfo().toException());
                    } else {
                        stage.abort();
                    }
                });
                queryStateMachine.transitionToFailed(stageFailureInfo.getFailureInfo().toException());
            }
        }
    });
    return Optional.of(distributedStagesScheduler);
}

Example 5 with ErrorCode

use of io.trino.spi.ErrorCode in project trino by trinodb.

the class TestQueryStateInfoResource method setUp.

@BeforeClass
public void setUp() {
    server = TestingTrinoServer.create();
    server.installPlugin(new TpchPlugin());
    server.createCatalog("tpch", "tpch");
    client = new JettyHttpClient();
    Request request1 = preparePost().setUri(uriBuilderFrom(server.getBaseUrl()).replacePath("/v1/statement").build()).setBodyGenerator(createStaticBodyGenerator(LONG_LASTING_QUERY, UTF_8)).setHeader(TRINO_HEADERS.requestUser(), "user1").build();
    queryResults = client.execute(request1, createJsonResponseHandler(QUERY_RESULTS_JSON_CODEC));
    client.execute(prepareGet().setUri(queryResults.getNextUri()).build(), createJsonResponseHandler(QUERY_RESULTS_JSON_CODEC));
    Request request2 = preparePost().setUri(uriBuilderFrom(server.getBaseUrl()).replacePath("/v1/statement").build()).setBodyGenerator(createStaticBodyGenerator(LONG_LASTING_QUERY, UTF_8)).setHeader(TRINO_HEADERS.requestUser(), "user2").build();
    QueryResults queryResults2 = client.execute(request2, createJsonResponseHandler(jsonCodec(QueryResults.class)));
    client.execute(prepareGet().setUri(queryResults2.getNextUri()).build(), createJsonResponseHandler(QUERY_RESULTS_JSON_CODEC));
    // queries are started in the background, so they may not all be immediately visible
    long start = System.nanoTime();
    while (Duration.nanosSince(start).compareTo(new Duration(5, MINUTES)) < 0) {
        List<BasicQueryInfo> queryInfos = client.execute(prepareGet().setUri(uriBuilderFrom(server.getBaseUrl()).replacePath("/v1/query").build()).setHeader(TRINO_HEADERS.requestUser(), "unknown").build(), createJsonResponseHandler(listJsonCodec(BasicQueryInfo.class)));
        if (queryInfos.size() == 2) {
            if (queryInfos.stream().allMatch(info -> info.getState() == RUNNING)) {
                break;
            }
            List<ErrorCode> errorCodes = queryInfos.stream().filter(info -> info.getState() == FAILED).map(BasicQueryInfo::getErrorCode).collect(toImmutableList());
            if (!errorCodes.isEmpty()) {
                fail("setup queries failed with: " + errorCodes);
            }
        }
    }
}