Examples with ConnectionException org.apache.nifi.cluster.ConnectionException used on opensource projects

Example 1 with ConnectionException

use of org.apache.nifi.cluster.ConnectionException in project nifi by apache.

the class StandardFlowService method connect.

private ConnectionResponse connect(final boolean retryOnCommsFailure, final boolean retryIndefinitely, final DataFlow dataFlow) throws ConnectionException {
    readLock.lock();
    try {
        logger.info("Connecting Node: " + nodeId);
        // create connection request message
        final ConnectionRequest request = new ConnectionRequest(nodeId, dataFlow);
        final ConnectionRequestMessage requestMsg = new ConnectionRequestMessage();
        requestMsg.setConnectionRequest(request);
        // send connection request to cluster manager
        /*
             * Try to get a current copy of the cluster's dataflow from the manager
             * for ten times, sleeping between attempts. Ten times should be
             * enough because the manager will register the node as connecting
             * and therefore, no other changes to the cluster flow can occur.
             *
             * However, the manager needs to obtain a current data flow within
             * maxAttempts * tryLaterSeconds or else the node will fail to startup.
             */
        final int maxAttempts = 10;
        ConnectionResponse response = null;
        for (int i = 0; i < maxAttempts || retryIndefinitely; i++) {
            try {
                response = senderListener.requestConnection(requestMsg).getConnectionResponse();
                if (response.shouldTryLater()) {
                    logger.info("Requested by cluster coordinator to retry connection in " + response.getTryLaterSeconds() + " seconds with explanation: " + response.getRejectionReason());
                    try {
                        Thread.sleep(response.getTryLaterSeconds() * 1000);
                    } catch (final InterruptedException ie) {
                        // we were interrupted, so finish quickly
                        Thread.currentThread().interrupt();
                        break;
                    }
                } else if (response.getRejectionReason() != null) {
                    logger.warn("Connection request was blocked by cluster coordinator with the explanation: " + response.getRejectionReason());
                    // set response to null and treat a firewall blockage the same as getting no response from manager
                    response = null;
                    break;
                } else {
                    // we received a successful connection response from manager
                    break;
                }
            } catch (final NoClusterCoordinatorException ncce) {
                logger.warn("There is currently no Cluster Coordinator. This often happens upon restart of NiFi when running an embedded ZooKeeper. Will register this node " + "to become the active Cluster Coordinator and will attempt to connect to cluster again");
                controller.registerForClusterCoordinator(true);
                try {
                    Thread.sleep(1000L);
                } catch (final InterruptedException ie) {
                    Thread.currentThread().interrupt();
                    break;
                }
            } catch (final Exception pe) {
                // could not create a socket and communicate with manager
                logger.warn("Failed to connect to cluster due to: " + pe);
                if (logger.isDebugEnabled()) {
                    logger.warn("", pe);
                }
                if (retryOnCommsFailure) {
                    try {
                        Thread.sleep(response == null ? 5000 : response.getTryLaterSeconds());
                    } catch (final InterruptedException ie) {
                        Thread.currentThread().interrupt();
                        break;
                    }
                } else {
                    break;
                }
            }
        }
        if (response == null) {
            // if response is null, then either we had IO problems or we were blocked by firewall or we couldn't determine manager's address
            return response;
        } else if (response.shouldTryLater()) {
            // if response indicates we should try later, then coordinator was unable to service our request. Just load local flow and move on.
            // when the cluster coordinator is able to service requests, this node's heartbeat will trigger the cluster coordinator to reach
            // out to this node and re-connect to the cluster.
            logger.info("Received a 'try again' response from Cluster Coordinator when attempting to connect to cluster with explanation '" + response.getRejectionReason() + "'. However, the maximum number of retries have already completed. Will load local flow and connect to the cluster when able.");
            return null;
        } else {
            // persist node uuid and index returned by NCM and return the response to the caller
            try {
                // Ensure that we have registered our 'cluster node configuration' state key
                final Map<String, String> map = Collections.singletonMap(NODE_UUID, response.getNodeIdentifier().getId());
                controller.getStateManagerProvider().getStateManager(CLUSTER_NODE_CONFIG).setState(map, Scope.LOCAL);
            } catch (final IOException ioe) {
                logger.warn("Received successful response from Cluster Manager but failed to persist state about the Node's Unique Identifier and the Node's Index. " + "This node may be assigned a different UUID when the node is restarted.", ioe);
            }
            return response;
        }
    } finally {
        readLock.unlock();
    }
}

Example 2 with ConnectionException

use of org.apache.nifi.cluster.ConnectionException in project nifi by apache.

the class StandardFlowService method loadFromConnectionResponse.

private void loadFromConnectionResponse(final ConnectionResponse response) throws ConnectionException {
    writeLock.lock();
    try {
        if (response.getNodeConnectionStatuses() != null) {
            clusterCoordinator.resetNodeStatuses(response.getNodeConnectionStatuses().stream().collect(Collectors.toMap(status -> status.getNodeIdentifier(), status -> status)));
        }
        // get the dataflow from the response
        final DataFlow dataFlow = response.getDataFlow();
        if (logger.isTraceEnabled()) {
            logger.trace("ResponseFlow = " + new String(dataFlow.getFlow(), StandardCharsets.UTF_8));
        }
        // load new controller state
        loadFromBytes(dataFlow, true);
        // set node ID on controller before we start heartbeating because heartbeat needs node ID
        nodeId = response.getNodeIdentifier();
        logger.info("Setting Flow Controller's Node ID: " + nodeId);
        controller.setNodeId(nodeId);
        clusterCoordinator.setLocalNodeIdentifier(nodeId);
        clusterCoordinator.setConnected(true);
        revisionManager.reset(response.getComponentRevisions().stream().map(rev -> rev.toRevision()).collect(Collectors.toList()));
        // mark the node as clustered
        controller.setClustered(true, response.getInstanceId());
        controller.setConnectionStatus(new NodeConnectionStatus(nodeId, NodeConnectionState.CONNECTED));
        // Initialize the controller after the flow is loaded so we don't take any actions on repos until everything is good
        initializeController();
        // start the processors as indicated by the dataflow
        controller.onFlowInitialized(autoResumeState);
        loadSnippets(dataFlow.getSnippets());
        controller.startHeartbeating();
    } catch (final UninheritableFlowException ufe) {
        throw new UninheritableFlowException(CONNECTION_EXCEPTION_MSG_PREFIX + "local flow is different than cluster flow.", ufe);
    } catch (final MissingBundleException mbe) {
        throw new MissingBundleException(CONNECTION_EXCEPTION_MSG_PREFIX + "cluster flow contains bundles that do not exist on the current node", mbe);
    } catch (final FlowSerializationException fse) {
        throw new ConnectionException(CONNECTION_EXCEPTION_MSG_PREFIX + "local or cluster flow is malformed.", fse);
    } catch (final FlowSynchronizationException fse) {
        throw new FlowSynchronizationException(CONNECTION_EXCEPTION_MSG_PREFIX + "local flow controller partially updated. " + "Administrator should disconnect node and review flow for corruption.", fse);
    } catch (final Exception ex) {
        throw new ConnectionException("Failed to connect node to cluster due to: " + ex, ex);
    } finally {
        writeLock.unlock();
    }
}