Search in sources :

Example 1 with PendingEventTooLargeException

use of com.linkedin.databus.core.PendingEventTooLargeException in project databus by linkedin.

the class RelayPullThread method doReadDataEvents.

protected void doReadDataEvents(ConnectionState curState) {
    boolean debugEnabled = _log.isDebugEnabled();
    boolean enqueueMessage = true;
    try {
        ChunkedBodyReadableByteChannel readChannel = curState.getReadChannel();
        Checkpoint cp = curState.getCheckpoint();
        curState.setRelayFellOff(false);
        String remoteErrorName = RemoteExceptionHandler.getExceptionName(readChannel);
        Throwable knownRemoteError = _remoteExceptionHandler.getException(readChannel);
        if (null != knownRemoteError && knownRemoteError instanceof ScnNotFoundException) {
            if (toTearConnAfterHandlingResponse()) {
                tearConnectionAndEnqueuePickServer();
                enqueueMessage = false;
            } else {
                curState.setRelayFellOff(true);
                if (_retriesOnFallOff.getRemainingRetriesNum() > 0) {
                    _log.error("Got SCNNotFoundException. Retry (" + _retriesOnFallOff.getRetriesNum() + ") out of " + _retriesOnFallOff.getConfig().getMaxRetryNum());
                    curState.switchToPickServer();
                } else {
                    enqueueMessage = onRelayFellOff(curState, cp, knownRemoteError);
                }
            }
        } else if (null != remoteErrorName) {
            if (toTearConnAfterHandlingResponse()) {
                tearConnectionAndEnqueuePickServer();
                enqueueMessage = false;
            } else {
                // remote processing error
                _log.error("read events error: " + RemoteExceptionHandler.getExceptionMessage(readChannel));
                curState.switchToStreamResponseError();
            }
        } else {
            /*DispatcherState dispatchState = curState.getDispatcherState();
          dispatchState.switchToDispatchEvents();
          _dispatcherThread.addNewStateBlocking(dispatchState);*/
            if (debugEnabled)
                _log.debug("Sending events to buffer");
            DbusEventsStatisticsCollector connCollector = _sourcesConn.getInboundEventsStatsCollector();
            if (curState.isSCNRegress()) {
                _log.info("SCN Regress requested !! Sending a SCN Regress Message to dispatcher. Curr Ckpt :" + curState.getCheckpoint());
                DbusEvent regressEvent = getEventFactory().createSCNRegressEvent(new SCNRegressMessage(curState.getCheckpoint()));
                writeEventToRelayDispatcher(curState, regressEvent, "SCN Regress Event from ckpt :" + curState.getCheckpoint());
                curState.setSCNRegress(false);
            }
            UnifiedClientStats unifiedClientStats = _sourcesConn.getUnifiedClientStats();
            if (unifiedClientStats != null) {
                // failsafe:  we're definitely not bootstrapping here
                unifiedClientStats.setBootstrappingState(false);
                sendHeartbeat(unifiedClientStats);
            }
            int eventsNum = curState.getDataEventsBuffer().readEvents(readChannel, curState.getListeners(), connCollector);
            boolean resetConnection = false;
            if (eventsNum > 0) {
                _timeSinceEventsSec = System.currentTimeMillis();
                cp.checkPoint();
            } else {
                // check how long it has been since we got some events
                if (_remoteExceptionHandler.getPendingEventSize(readChannel) > curState.getDataEventsBuffer().getMaxReadBufferCapacity()) {
                    // The relay had a pending event that we can never accommodate. This is fatal error.
                    String err = "ReadBuffer max capacity(" + curState.getDataEventsBuffer().getMaxReadBufferCapacity() + ") is less than event size(" + _remoteExceptionHandler.getPendingEventSize(readChannel) + "). Increase databus.client.connectionDefaults.eventBuffer.maxEventSize and restart.";
                    _log.fatal(err);
                    enqueueMessage(LifecycleMessage.createSuspendOnErroMessage(new PendingEventTooLargeException(err)));
                    return;
                } else {
                    if (_noEventsConnectionResetTimeSec > 0) {
                        // unless the feature is disabled (<=0)
                        resetConnection = (System.currentTimeMillis() - _timeSinceEventsSec) / 1000 > _noEventsConnectionResetTimeSec;
                        if (resetConnection) {
                            _timeSinceEventsSec = System.currentTimeMillis();
                            _log.warn("about to reset connection to relay " + curState.getServerInetAddress() + ", because there were no events for " + _noEventsConnectionResetTimeSec + "secs");
                        }
                    }
                }
            }
            if (debugEnabled)
                _log.debug("Events read: " + eventsNum);
            // if it has been too long since we got non-empty responses - the relay may be stuck, try to reconnect
            if (toTearConnAfterHandlingResponse() || resetConnection) {
                tearConnectionAndEnqueuePickServer();
                enqueueMessage = false;
            } else {
                curState.switchToStreamResponseDone();
                resetServerRetries();
            }
        }
        if (enqueueMessage)
            enqueueMessage(curState);
    } catch (InterruptedException ie) {
        _log.warn("interrupted", ie);
        curState.switchToStreamResponseError();
        enqueueMessage(curState);
    } catch (InvalidEventException e) {
        _log.error("error reading events from server:" + e, e);
        curState.switchToStreamResponseError();
        enqueueMessage(curState);
    } catch (RuntimeException e) {
        _log.error("runtime error reading events from server: " + e, e);
        curState.switchToStreamResponseError();
        enqueueMessage(curState);
    }
}
Also used : UnifiedClientStats(com.linkedin.databus.client.pub.mbean.UnifiedClientStats) PendingEventTooLargeException(com.linkedin.databus.core.PendingEventTooLargeException) DbusEvent(com.linkedin.databus.core.DbusEvent) DbusEventsStatisticsCollector(com.linkedin.databus.core.monitoring.mbean.DbusEventsStatisticsCollector) Checkpoint(com.linkedin.databus.core.Checkpoint) ScnNotFoundException(com.linkedin.databus.core.ScnNotFoundException) SCNRegressMessage(com.linkedin.databus.core.SCNRegressMessage) InvalidEventException(com.linkedin.databus.core.InvalidEventException)

Example 2 with PendingEventTooLargeException

use of com.linkedin.databus.core.PendingEventTooLargeException in project databus by linkedin.

the class BootstrapPullThread method doReadBootstrapEvents.

protected void doReadBootstrapEvents(ConnectionState curState) {
    boolean success = true;
    boolean debugEnabled = _log.isDebugEnabled();
    boolean enqueueMessage = true;
    try {
        Checkpoint cp = curState.getCheckpoint();
        DbusEventBuffer eventBuffer = curState.getDataEventsBuffer();
        if (debugEnabled)
            _log.debug("Sending bootstrap events to buffer");
        // eventBuffer.startEvents();
        DbusEventInternalReadable cpEvent = getEventFactory().createCheckpointEvent(cp);
        byte[] cpEventBytes = new byte[cpEvent.size()];
        if (debugEnabled) {
            _log.debug("checkpoint event size: " + cpEventBytes.length);
            _log.debug("checkpoint event:" + cpEvent.toString());
        }
        cpEvent.getRawBytes().get(cpEventBytes);
        ByteArrayInputStream cpIs = new ByteArrayInputStream(cpEventBytes);
        ReadableByteChannel cpRbc = Channels.newChannel(cpIs);
        UnifiedClientStats unifiedClientStats = _sourcesConn.getUnifiedClientStats();
        sendHeartbeat(unifiedClientStats);
        int ecnt = eventBuffer.readEvents(cpRbc);
        success = (ecnt > 0);
        if (!success) {
            _log.error("Unable to write bootstrap phase marker");
        } else {
            ChunkedBodyReadableByteChannel readChannel = curState.getReadChannel();
            String remoteErrorName = RemoteExceptionHandler.getExceptionName(readChannel);
            Throwable remoteError = _remoteExceptionHandler.getException(readChannel);
            if (null != remoteError && remoteError instanceof BootstrapDatabaseTooOldException) {
                _log.error("Bootstrap database is too old!");
                _remoteExceptionHandler.handleException(remoteError);
                curState.switchToStreamResponseError();
            } else if (null != remoteErrorName) {
                // remote processing error
                _log.error("read events error: " + RemoteExceptionHandler.getExceptionMessage(readChannel));
                curState.switchToStreamResponseError();
            } else {
                sendHeartbeat(unifiedClientStats);
                int eventsNum = eventBuffer.readEvents(readChannel, curState.getListeners(), _sourcesConn.getBootstrapEventsStatsCollector());
                if (eventsNum == 0 && _remoteExceptionHandler.getPendingEventSize(readChannel) > eventBuffer.getMaxReadBufferCapacity()) {
                    String err = "ReadBuffer max capacity(" + eventBuffer.getMaxReadBufferCapacity() + ") is less than event size(" + _remoteExceptionHandler.getPendingEventSize(readChannel) + "). Increase databus.client.connectionDefaults.bstEventBuffer.maxEventSize and restart.";
                    _log.fatal(err);
                    enqueueMessage(LifecycleMessage.createSuspendOnErroMessage(new PendingEventTooLargeException(err)));
                    return;
                } else {
                    resetServerRetries();
                    if (debugEnabled)
                        _log.debug("Sending events to buffer");
                    numEventsInCurrentState += eventsNum;
                    _log.info("Bootstrap events read so far: " + numEventsInCurrentState);
                    String status = readChannel.getMetadata("PhaseCompleted");
                    final BootstrapCheckpointHandler ckptHandler = curState.getBstCheckpointHandler();
                    if (status != null) {
                        // set status in checkpoint to indicate that we are done with the current source
                        if (cp.getConsumptionMode() == DbusClientMode.BOOTSTRAP_CATCHUP) {
                            ckptHandler.finalizeCatchupPhase(cp);
                        } else if (cp.getConsumptionMode() == DbusClientMode.BOOTSTRAP_SNAPSHOT) {
                            ckptHandler.finalizeSnapshotPhase(cp);
                        } else {
                            throw new RuntimeException("Invalid bootstrap phase: " + cp.getConsumptionMode());
                        }
                        _log.info("Bootstrap events read :" + numEventsInCurrentState + " during phase:" + cp.getConsumptionMode() + " [" + cp.getBootstrapSnapshotSourceIndex() + "," + cp.getBootstrapCatchupSourceIndex() + "]");
                        numEventsInCurrentState = 0;
                    } else {
                        // question: how is snapshotOffset maintained in ckpt
                        if (eventsNum > 0) {
                            cp.bootstrapCheckPoint();
                        }
                    }
                    curState.switchToStreamResponseDone();
                }
            }
        }
    } catch (InterruptedException ie) {
        _log.error("interupted", ie);
        success = false;
    } catch (InvalidEventException e) {
        _log.error("error reading events from server: " + e.getMessage(), e);
        success = false;
    } catch (RuntimeException e) {
        _log.error("runtime error reading events from server: " + e.getMessage(), e);
        success = false;
    }
    if (toTearConnAfterHandlingResponse()) {
        tearConnectionAndEnqueuePickServer();
        enqueueMessage = false;
    } else if (!success) {
        curState.switchToPickServer();
    }
    if (enqueueMessage)
        enqueueMessage(curState);
}
Also used : ReadableByteChannel(java.nio.channels.ReadableByteChannel) UnifiedClientStats(com.linkedin.databus.client.pub.mbean.UnifiedClientStats) PendingEventTooLargeException(com.linkedin.databus.core.PendingEventTooLargeException) DbusEventInternalReadable(com.linkedin.databus.core.DbusEventInternalReadable) Checkpoint(com.linkedin.databus.core.Checkpoint) BootstrapCheckpointHandler(com.linkedin.databus.core.BootstrapCheckpointHandler) DbusEventBuffer(com.linkedin.databus.core.DbusEventBuffer) BootstrapDatabaseTooOldException(com.linkedin.databus2.core.container.request.BootstrapDatabaseTooOldException) Checkpoint(com.linkedin.databus.core.Checkpoint) ByteArrayInputStream(java.io.ByteArrayInputStream) InvalidEventException(com.linkedin.databus.core.InvalidEventException)

Aggregations

UnifiedClientStats (com.linkedin.databus.client.pub.mbean.UnifiedClientStats)2 Checkpoint (com.linkedin.databus.core.Checkpoint)2 InvalidEventException (com.linkedin.databus.core.InvalidEventException)2 PendingEventTooLargeException (com.linkedin.databus.core.PendingEventTooLargeException)2 BootstrapCheckpointHandler (com.linkedin.databus.core.BootstrapCheckpointHandler)1 DbusEvent (com.linkedin.databus.core.DbusEvent)1 DbusEventBuffer (com.linkedin.databus.core.DbusEventBuffer)1 DbusEventInternalReadable (com.linkedin.databus.core.DbusEventInternalReadable)1 SCNRegressMessage (com.linkedin.databus.core.SCNRegressMessage)1 ScnNotFoundException (com.linkedin.databus.core.ScnNotFoundException)1 DbusEventsStatisticsCollector (com.linkedin.databus.core.monitoring.mbean.DbusEventsStatisticsCollector)1 BootstrapDatabaseTooOldException (com.linkedin.databus2.core.container.request.BootstrapDatabaseTooOldException)1 ByteArrayInputStream (java.io.ByteArrayInputStream)1 ReadableByteChannel (java.nio.channels.ReadableByteChannel)1