Examples with NodeControllerState org.apache.hyracks.control.cc.NodeControllerState used on opensource projects

Example 6 with NodeControllerState

use of org.apache.hyracks.control.cc.NodeControllerState in project asterixdb by apache.

the class CliDeployBinaryWork method doRun.

@Override
public void doRun() {
    try {
        if (deploymentId == null) {
            deploymentId = new DeploymentId(UUID.randomUUID().toString());
        }
        /**
             * Deploy for the cluster controller
             */
        DeploymentUtils.deploy(deploymentId, binaryURLs, ccs.getContext().getJobSerializerDeserializerContainer(), ccs.getServerContext(), false);
        /**
             * Deploy for the node controllers
             */
        INodeManager nodeManager = ccs.getNodeManager();
        Collection<String> nodeIds = nodeManager.getAllNodeIds();
        final DeploymentRun dRun = new DeploymentRun(nodeIds);
        /** The following call prevents a user to deploy with the same deployment id simultaneously. */
        ccs.addDeploymentRun(deploymentId, dRun);
        /***
             * deploy binaries to each node controller
             */
        for (NodeControllerState ncs : nodeManager.getAllNodeControllerStates()) {
            ncs.getNodeController().deployBinary(deploymentId, binaryURLs);
        }
        ccs.getExecutor().execute(new Runnable() {

            @Override
            public void run() {
                try {
                    /**
                         * wait for completion
                         */
                    dRun.waitForCompletion();
                    ccs.removeDeploymentRun(deploymentId);
                    callback.setValue(deploymentId);
                } catch (Exception e) {
                    callback.setException(e);
                }
            }
        });
    } catch (Exception e) {
        callback.setException(e);
    }
}

Example 7 with NodeControllerState

use of org.apache.hyracks.control.cc.NodeControllerState in project asterixdb by apache.

the class CliUnDeployBinaryWork method doRun.

@Override
public void doRun() {
    try {
        if (deploymentId == null) {
            deploymentId = new DeploymentId(UUID.randomUUID().toString());
        }
        /**
             * Deploy for the cluster controller
             */
        DeploymentUtils.undeploy(deploymentId, ccs.getContext().getJobSerializerDeserializerContainer(), ccs.getServerContext());
        /**
             * Deploy for the node controllers
             */
        INodeManager nodeManager = ccs.getNodeManager();
        Collection<String> nodeIds = nodeManager.getAllNodeIds();
        final DeploymentRun dRun = new DeploymentRun(nodeIds);
        /** The following call prevents a user to undeploy with the same deployment id simultaneously. */
        ccs.addDeploymentRun(deploymentId, dRun);
        /***
             * deploy binaries to each node controller
             */
        for (NodeControllerState ncs : nodeManager.getAllNodeControllerStates()) {
            ncs.getNodeController().undeployBinary(deploymentId);
        }
        ccs.getExecutor().execute(new Runnable() {

            @Override
            public void run() {
                try {
                    /**
                         * wait for completion
                         */
                    dRun.waitForCompletion();
                    ccs.removeDeploymentRun(deploymentId);
                    callback.setValue(null);
                } catch (Exception e) {
                    callback.setException(e);
                }
            }
        });
    } catch (Exception e) {
        throw new RuntimeException(e);
    }
}

Example 8 with NodeControllerState

use of org.apache.hyracks.control.cc.NodeControllerState in project asterixdb by apache.

the class DestroyJobWork method doRun.

@Override
protected void doRun() throws Exception {
    try {
        ccs.getPreDistributedJobStore().removeDistributedJobDescriptor(jobId);
        INodeManager nodeManager = ccs.getNodeManager();
        for (NodeControllerState node : nodeManager.getAllNodeControllerStates()) {
            node.getNodeController().destroyJob(jobId);
        }
        callback.setValue(jobId);
    } catch (Exception e) {
        callback.setException(e);
    }
}

Example 9 with NodeControllerState

use of org.apache.hyracks.control.cc.NodeControllerState in project asterixdb by apache.

the class JobExecutor method abortTaskCluster.

private void abortTaskCluster(TaskClusterAttempt tcAttempt, TaskClusterAttempt.TaskClusterStatus failedOrAbortedStatus) {
    LOGGER.fine("Aborting task cluster: " + tcAttempt.getAttempt());
    Set<TaskAttemptId> abortTaskIds = new HashSet<>();
    Map<String, List<TaskAttemptId>> abortTaskAttemptMap = new HashMap<>();
    for (TaskAttempt ta : tcAttempt.getTaskAttempts().values()) {
        TaskAttemptId taId = ta.getTaskAttemptId();
        TaskAttempt.TaskStatus status = ta.getStatus();
        abortTaskIds.add(taId);
        LOGGER.fine("Checking " + taId + ": " + ta.getStatus());
        if (status == TaskAttempt.TaskStatus.RUNNING || status == TaskAttempt.TaskStatus.COMPLETED) {
            ta.setStatus(TaskAttempt.TaskStatus.ABORTED, null);
            ta.setEndTime(System.currentTimeMillis());
            List<TaskAttemptId> abortTaskAttempts = abortTaskAttemptMap.get(ta.getNodeId());
            if (status == TaskAttempt.TaskStatus.RUNNING && abortTaskAttempts == null) {
                abortTaskAttempts = new ArrayList<>();
                abortTaskAttemptMap.put(ta.getNodeId(), abortTaskAttempts);
            }
            if (status == TaskAttempt.TaskStatus.RUNNING) {
                abortTaskAttempts.add(taId);
            }
        }
    }
    final JobId jobId = jobRun.getJobId();
    LOGGER.fine("Abort map for job: " + jobId + ": " + abortTaskAttemptMap);
    INodeManager nodeManager = ccs.getNodeManager();
    for (Map.Entry<String, List<TaskAttemptId>> entry : abortTaskAttemptMap.entrySet()) {
        final NodeControllerState node = nodeManager.getNodeControllerState(entry.getKey());
        final List<TaskAttemptId> abortTaskAttempts = entry.getValue();
        if (node != null) {
            if (LOGGER.isLoggable(Level.FINE)) {
                LOGGER.fine("Aborting: " + abortTaskAttempts + " at " + entry.getKey());
            }
            try {
                node.getNodeController().abortTasks(jobId, abortTaskAttempts);
            } catch (Exception e) {
                LOGGER.log(Level.SEVERE, e.getMessage(), e);
            }
        }
    }
    inProgressTaskClusters.remove(tcAttempt.getTaskCluster());
    TaskCluster tc = tcAttempt.getTaskCluster();
    PartitionMatchMaker pmm = jobRun.getPartitionMatchMaker();
    pmm.removeUncommittedPartitions(tc.getProducedPartitions(), abortTaskIds);
    pmm.removePartitionRequests(tc.getRequiredPartitions(), abortTaskIds);
    tcAttempt.setStatus(failedOrAbortedStatus);
    tcAttempt.setEndTime(System.currentTimeMillis());
}

Example 10 with NodeControllerState

use of org.apache.hyracks.control.cc.NodeControllerState in project asterixdb by apache.

the class NodeManager method removeDeadNodes.

@Override
public Pair<Collection<String>, Collection<JobId>> removeDeadNodes() throws HyracksException {
    Set<String> deadNodes = new HashSet<>();
    Set<JobId> affectedJobIds = new HashSet<>();
    Iterator<Map.Entry<String, NodeControllerState>> nodeIterator = nodeRegistry.entrySet().iterator();
    while (nodeIterator.hasNext()) {
        Map.Entry<String, NodeControllerState> entry = nodeIterator.next();
        String nodeId = entry.getKey();
        NodeControllerState state = entry.getValue();
        if (state.incrementLastHeartbeatDuration() >= ccConfig.getHeartbeatMaxMisses()) {
            deadNodes.add(nodeId);
            affectedJobIds.addAll(state.getActiveJobIds());
            // Removes the node from node map.
            nodeIterator.remove();
            // Removes the node from IP map.
            removeNodeFromIpAddressMap(nodeId, state);
            // Updates the cluster capacity.
            resourceManager.update(nodeId, new NodeCapacity(0L, 0));
            LOGGER.info(entry.getKey() + " considered dead");
        }
    }
    return Pair.of(deadNodes, affectedJobIds);
}