Search in sources :

Example 66 with Client

use of org.ow2.proactive.resourcemanager.authentication.Client in project scheduling by ow2-proactive.

the class RMCore method releaseNodes.

/**
 * {@inheritDoc}
 */
public BooleanWrapper releaseNodes(NodeSet nodes) {
    if (nodes.getExtraNodes() != null) {
        // do not forget to release extra nodes
        nodes.addAll(nodes.getExtraNodes());
    }
    // exception to throw in case of problems
    RuntimeException exception = null;
    NodeSet nodesReleased = new NodeSet();
    NodeSet nodesFailedToRelease = new NodeSet();
    for (Node node : nodes) {
        String nodeURL = null;
        try {
            nodeURL = node.getNodeInformation().getURL();
            logger.debug("Releasing node " + nodeURL);
        } catch (RuntimeException e) {
            logger.debug("A Runtime exception occurred while obtaining information on the node," + "the node must be down (it will be detected later)", e);
            // node is down, will be detected by pinger
            exception = new IllegalStateException(e.getMessage(), e);
            nodesFailedToRelease.add(node);
        }
        // verify whether the node has not been removed from the RM
        if (this.allNodes.containsKey(nodeURL)) {
            RMNode rmnode = this.getNodebyUrl(nodeURL);
            // free
            if (rmnode.isFree()) {
                logger.warn("Client " + caller + " tries to release the already free node " + nodeURL);
                nodesFailedToRelease.add(node);
            } else if (rmnode.isDown()) {
                logger.warn("Node was down, it cannot be released");
                nodesFailedToRelease.add(node);
            } else {
                Set<? extends IdentityPrincipal> userPrincipal = rmnode.getOwner().getSubject().getPrincipals(UserNamePrincipal.class);
                Permission ownerPermission = new PrincipalPermission(rmnode.getOwner().getName(), userPrincipal);
                try {
                    caller.checkPermission(ownerPermission, caller + " is not authorized to free node " + node.getNodeInformation().getURL(), new RMCoreAllPermission(), new NodeUserAllPermission(), new NSAdminPermission());
                    if (rmnode.isToRemove()) {
                        removeNodeFromCoreAndSource(rmnode, caller);
                        nodesReleased.add(node);
                        if (delayedNodeSourceRemovalEvents.containsKey(rmnode.getNodeSourceName()) && nodeSourceCanBeRemoved(rmnode.getNodeSourceName())) {
                            logger.debug(NODE_SOURCE_STRING + rmnode.getNodeSourceName() + " is eligible to remove.");
                            final Entry<RMNodeSourceEvent, NodeSource> remove = delayedNodeSourceRemovalEvents.remove(rmnode.getNodeSourceName());
                            final RMNodeSourceEvent removedEvent = remove.getKey();
                            final NodeSource nodeSource = remove.getValue();
                            logger.info(NODE_SOURCE_STRING + rmnode.getNodeSourceName() + HAS_BEEN_SUCCESSFULLY + removedEvent.getEventType().getDescription());
                            this.monitoring.nodeSourceEvent(removedEvent);
                            nodeSource.shutdown(this.caller);
                        } else if (delayedNodeSourceUndeploying.containsKey(rmnode.getNodeSourceName()) && nodeSourceCanBeRemoved(rmnode.getNodeSourceName())) {
                            logger.debug(NODE_SOURCE_STRING + rmnode.getNodeSourceName() + " is eligible to undeploy.");
                            final NodeSource nodeSource = delayedNodeSourceUndeploying.remove(rmnode.getNodeSourceName());
                            logger.info(NODE_SOURCE_STRING + rmnode.getNodeSourceName() + HAS_BEEN_SUCCESSFULLY + "undeployed.");
                            nodeSource.shutdown(this.caller);
                        }
                    } else {
                        internalSetFree(rmnode);
                        nodesReleased.add(node);
                    }
                } catch (SecurityException ex) {
                    logger.error(ex.getMessage(), ex);
                    nodesFailedToRelease.add(node);
                    exception = ex;
                }
            }
        } else {
            logger.warn("Cannot release unknown node " + nodeURL);
            nodesFailedToRelease.add(node);
            exception = new IllegalArgumentException("Cannot release unknown node " + nodeURL);
        }
    }
    logger.info("Nodes released : " + nodesReleased);
    if (!nodesFailedToRelease.isEmpty()) {
        logger.warn("Nodes failed to release : " + nodesFailedToRelease);
    }
    if (exception != null) {
        // throwing the latest exception we had
        throw exception;
    }
    return new BooleanWrapper(true);
}
Also used : NodeSet(org.ow2.proactive.utils.NodeSet) RMNodeSourceEvent(org.ow2.proactive.resourcemanager.common.event.RMNodeSourceEvent) NodeSet(org.ow2.proactive.utils.NodeSet) Set(java.util.Set) ImmutableSet(com.google.common.collect.ImmutableSet) HashSet(java.util.HashSet) RMDeployingNode(org.ow2.proactive.resourcemanager.rmnode.RMDeployingNode) Node(org.objectweb.proactive.core.node.Node) RMNode(org.ow2.proactive.resourcemanager.rmnode.RMNode) PrincipalPermission(org.ow2.proactive.permissions.PrincipalPermission) NodeUserAllPermission(org.ow2.proactive.permissions.NodeUserAllPermission) UserNamePrincipal(org.ow2.proactive.authentication.principals.UserNamePrincipal) RMCoreAllPermission(org.ow2.proactive.permissions.RMCoreAllPermission) NSAdminPermission(org.ow2.proactive.permissions.NSAdminPermission) NodeSource(org.ow2.proactive.resourcemanager.nodesource.NodeSource) BooleanWrapper(org.objectweb.proactive.core.util.wrapper.BooleanWrapper) RMNode(org.ow2.proactive.resourcemanager.rmnode.RMNode) Entry(java.util.Map.Entry) Permission(java.security.Permission) NSAdminPermission(org.ow2.proactive.permissions.NSAdminPermission) NodeUserAllPermission(org.ow2.proactive.permissions.NodeUserAllPermission) MethodCallPermission(org.ow2.proactive.permissions.MethodCallPermission) RMCoreAllPermission(org.ow2.proactive.permissions.RMCoreAllPermission) PrincipalPermission(org.ow2.proactive.permissions.PrincipalPermission) IdentityPrincipal(org.ow2.proactive.authentication.principals.IdentityPrincipal)

Example 67 with Client

use of org.ow2.proactive.resourcemanager.authentication.Client in project scheduling by ow2-proactive.

the class RMCore method internalSetToRemove.

/**
 * Mark node to be removed after releasing.
 *
 * @param rmNode node to be removed after node is released.
 */
void internalSetToRemove(final RMNode rmNode, Client initiator) {
    // If the node is already marked to be removed, so no need to go further
    if (rmNode.isToRemove()) {
        return;
    }
    if (logger.isDebugEnabled()) {
        logger.debug("Prepare to removing the node " + rmNode.getNodeURL());
    }
    // Get the previous state of the node needed for the event
    final NodeState previousNodeState = rmNode.getState();
    rmNode.setToRemove();
    persistUpdatedRMNodeIfRecoveryEnabled(rmNode);
    // create the event
    this.registerAndEmitNodeEvent(rmNode.createNodeEvent(RMEventType.NODE_STATE_CHANGED, previousNodeState, initiator.getName()));
}
Also used : NodeState(org.ow2.proactive.resourcemanager.common.NodeState)

Example 68 with Client

use of org.ow2.proactive.resourcemanager.authentication.Client in project scheduling by ow2-proactive.

the class RMCore method internalSetFree.

/**
 * Change the state of the node to free, after a Task has been completed by the specified node.
 * The node passed as parameter is moved to the list of nodes which are eligible for scheduling if it is not locked.
 * In all cases, an event informing the node state's change is propagated to RMMonitoring.
 *
 * @param rmNode node to set free.
 * @return true if the node successfully set as free, false if it was down before.
 */
@VisibleForTesting
BooleanWrapper internalSetFree(final RMNode rmNode) {
    if (logger.isDebugEnabled()) {
        logger.debug("Current node state " + rmNode.getState() + " " + rmNode.getNodeURL());
        logger.debug("Setting node state to free " + rmNode.getNodeURL());
    }
    // If the node is already free no need to go further
    if (rmNode.isFree()) {
        return new BooleanWrapper(true);
    }
    // Get the previous state of the node needed for the event
    final NodeState previousNodeState = rmNode.getState();
    Client client = rmNode.getOwner();
    if (client == null) {
        // node has been just configured, so the user initiated this action is the node provider
        client = rmNode.getProvider();
    }
    // resetting owner here
    rmNode.setFree();
    // an eligible node is a node that is free and not locked
    if (!rmNode.isLocked()) {
        this.eligibleNodes.add(rmNode);
    }
    persistUpdatedRMNodeIfRecoveryEnabled(rmNode);
    this.registerAndEmitNodeEvent(rmNode.createNodeEvent(RMEventType.NODE_STATE_CHANGED, previousNodeState, client.getName()));
    return new BooleanWrapper(true);
}
Also used : BooleanWrapper(org.objectweb.proactive.core.util.wrapper.BooleanWrapper) NodeState(org.ow2.proactive.resourcemanager.common.NodeState) Client(org.ow2.proactive.resourcemanager.authentication.Client) VisibleForTesting(com.google.common.annotations.VisibleForTesting)

Example 69 with Client

use of org.ow2.proactive.resourcemanager.authentication.Client in project scheduling by ow2-proactive.

the class RMCore method executeScript.

/**
 * {@inheritDoc}
 */
@ImmediateService
public <T> List<ScriptResult<T>> executeScript(Script<T> script, String targetType, Set<String> targets) {
    Client client;
    try {
        client = checkPermissionAndGetClientIsSuccessful();
    } catch (Exception e) {
        logger.error("Error while checking permission to execute node script", e);
        return Collections.singletonList(new ScriptResult<>(new ScriptException(e)));
    }
    // Depending on the target type, select nodes for script execution
    final TargetType tType = TargetType.valueOf(targetType);
    final HashSet<RMNode> selectedRMNodes = new HashSet<>();
    switch(tType) {
        case NODESOURCE_NAME:
            // If target is a nodesource name select one node, not busy if possible
            for (String target : targets) {
                NodeSource nodeSource = this.deployedNodeSources.get(target);
                if (nodeSource != null) {
                    Set<String> scriptExecutionHostNames = new HashSet<>();
                    for (RMNode candidateNode : this.allNodes.values()) {
                        String candidateNodeHostName = candidateNode.getHostName();
                        if (candidateNode.getNodeSource().equals(nodeSource) && !scriptExecutionHostNames.contains(candidateNodeHostName)) {
                            scriptExecutionHostNames.add(candidateNodeHostName);
                            this.selectCandidateNode(selectedRMNodes, candidateNode, client);
                        }
                    }
                }
            }
            break;
        case NODE_URL:
            // If target is node url select the node
            for (String target : targets) {
                RMNode candidateNode = this.allNodes.get(target);
                if (candidateNode != null) {
                    this.selectCandidateNode(selectedRMNodes, candidateNode, client);
                }
            }
            break;
        case HOSTNAME:
            // If target is hostname select first node from that host
            for (String target : targets) {
                for (RMNode node : this.allNodes.values()) {
                    if (node.getHostName().equals(target)) {
                        this.selectCandidateNode(selectedRMNodes, node, client);
                        break;
                    }
                }
            }
            break;
        default:
            throw new IllegalArgumentException("Unable to execute node script, unknown target type: " + targetType);
    }
    // Return a ProActive future on the list of results
    return this.selectionManager.executeScript(script, selectedRMNodes, null);
// To avoid blocking rmcore ao the call is delegated to the selection
// manager ao and each node is unlocked as soon as the script has
// finished it's execution.
}
Also used : ScriptResult(org.ow2.proactive.scripting.ScriptResult) ScriptException(org.ow2.proactive.scripting.ScriptException) NodeSource(org.ow2.proactive.resourcemanager.nodesource.NodeSource) RMNode(org.ow2.proactive.resourcemanager.rmnode.RMNode) TargetType(org.ow2.proactive.resourcemanager.utils.TargetType) Client(org.ow2.proactive.resourcemanager.authentication.Client) RMException(org.ow2.proactive.resourcemanager.exception.RMException) NodeException(org.objectweb.proactive.core.node.NodeException) ScriptException(org.ow2.proactive.scripting.ScriptException) PluginNotFoundException(org.ow2.proactive.resourcemanager.nodesource.PluginNotFoundException) TopologyException(org.ow2.proactive.resourcemanager.frontend.topology.TopologyException) ActiveObjectCreationException(org.objectweb.proactive.ActiveObjectCreationException) URISyntaxException(java.net.URISyntaxException) ThreadDumpNotAccessibleException(org.ow2.proactive.resourcemanager.rmnode.ThreadDumpNotAccessibleException) AddingNodesException(org.ow2.proactive.resourcemanager.exception.AddingNodesException) NotConnectedException(org.ow2.proactive.resourcemanager.exception.NotConnectedException) ProActiveException(org.objectweb.proactive.core.ProActiveException) HashSet(java.util.HashSet) ImmediateService(org.objectweb.proactive.annotation.ImmediateService)

Example 70 with Client

use of org.ow2.proactive.resourcemanager.authentication.Client in project scheduling by ow2-proactive.

the class SSHInfrastructureV2 method startNodeImpl.

/**
 * Internal node acquisition method
 * <p>
 * Starts a PA runtime on remote host using SSH, register it manually in the
 * nodesource.
 *
 * @param hostTracker The host on which one the node will be started
 * @param nbNodes number of nodes to deploy
 * @param depNodeURLs list of deploying or lost nodes urls created
 * @throws RMException
 *             acquisition failed
 */
public void startNodeImpl(final HostTracker hostTracker, final int nbNodes, final List<String> depNodeURLs) throws RMException {
    String fs = getTargetOSObj().fs;
    // we set the java security policy file
    ArrayList<String> sb = new ArrayList<>();
    final boolean containsSpace = schedulingPath.contains(" ");
    if (!deploymentMode.equals("useNodeJarStartupScript")) {
        if (containsSpace) {
            sb.add("-Dproactive.home=\"" + schedulingPath + "\"");
        } else {
            sb.add("-Dproactive.home=" + schedulingPath);
        }
    }
    String securitycmd = CentralPAPropertyRepository.JAVA_SECURITY_POLICY.getCmdLine();
    if (!this.javaOptions.contains(securitycmd) && !deploymentMode.equals("useNodeJarStartupScript")) {
        if (containsSpace) {
            securitycmd += "\"";
        }
        securitycmd += this.schedulingPath + fs + "config" + fs;
        securitycmd += "security.java.policy-client";
        if (containsSpace) {
            securitycmd += "\"";
        }
        sb.add(securitycmd);
    }
    // we set the log4j configuration file
    String log4jcmd = CentralPAPropertyRepository.LOG4J.getCmdLine();
    if (!this.javaOptions.contains(log4jcmd) && !deploymentMode.equals("useNodeJarStartupScript")) {
        // log4j only understands urls
        if (containsSpace) {
            log4jcmd += "\"";
        }
        log4jcmd += "file:";
        if (!this.schedulingPath.startsWith("/")) {
            log4jcmd += "/";
        }
        log4jcmd += this.schedulingPath.replace("\\", "/");
        log4jcmd += "/config/log/node.properties";
        if (containsSpace) {
            log4jcmd += "\"";
        }
        sb.add(log4jcmd);
    }
    // we add extra java/PA configuration
    if (this.javaOptions != null && !this.javaOptions.trim().isEmpty()) {
        sb.add(this.javaOptions.trim());
    }
    CommandLineBuilder clb = super.getDefaultCommandLineBuilder(getTargetOSObj());
    final boolean deployNodesInDetachedMode = PAResourceManagerProperties.RM_NODES_RECOVERY.getValueAsBoolean() || PAResourceManagerProperties.RM_PRESERVE_NODES_ON_SHUTDOWN.getValueAsBoolean();
    if (deployNodesInDetachedMode) {
        // if we do not want to kill the nodes when the RM exits or
        // restarts, then we should launch the nodes in background and
        // ignore the RM termination signal
        clb.setDetached();
    }
    clb.setJavaPath(this.javaPath);
    clb.setRmHome(this.schedulingPath);
    clb.setPaProperties(sb);
    final String nodeName = nodeNameBuilder.generateNodeName(hostTracker);
    clb.setNodeName(nodeName);
    clb.setNumberOfNodes(nbNodes);
    // set the stratup script retrieved from NodeCommandLine.properties
    if (!this.deploymentMode.equals("autoGenerated")) {
        clb.setDeploymentMode(deploymentMode);
        clb.setStartupScript((deploymentMode.equals("useStartupScript") ? startupScriptStandard : startupScriptWithNodeJarDownload));
    }
    if (this.deploymentMode.equals("useNodeJarStartupScript")) {
        clb.setNodeJarUrl(nodeJarUrl);
    }
    // finally, the credential's value
    String credString;
    try {
        Client currentClient = super.nodeSource.getAdministrator();
        credString = new String(currentClient.getCredentials().getBase64());
    } catch (KeyException e) {
        throw new RMException("Could not get base64 credentials", e);
    }
    clb.setCredentialsValueAndNullOthers(credString);
    // add an expected node. every unexpected node will be discarded
    String cmdLine;
    String obfuscatedCmdLine;
    try {
        cmdLine = clb.buildCommandLine(true);
        obfuscatedCmdLine = clb.buildCommandLine(false);
    } catch (IOException e) {
        throw new RMException("Cannot build the " + RMNodeStarter.class.getSimpleName() + "'s command line.", e);
    }
    // one escape the command to make it runnable through ssh
    if (cmdLine.contains("\"")) {
        cmdLine = cmdLine.replaceAll("\"", "\\\\\"");
    }
    final String finalCmdLine = cmdLine;
    // The final addDeployingNode() method will initiate a timeout that
    // will declare node as lost and set the description of the failure
    // with a simplistic message, since there is no way to override this
    // mechanism we consider only 90% of timeout to set custom description
    // in case of failure and still allow global timeout
    final int shorterTimeout = Math.round((90 * super.nodeTimeOut) / 100);
    JSch jsch = new JSch();
    final String msg = "deploy on " + hostTracker.getResolvedAddress();
    final List<String> createdNodeNames = RMNodeStarter.getWorkersNodeNames(nodeName, nbNodes);
    depNodeURLs.addAll(addMultipleDeployingNodes(createdNodeNames, obfuscatedCmdLine, msg, super.nodeTimeOut));
    addTimeouts(depNodeURLs);
    Session session;
    try {
        // Create ssh session to the hostname
        session = jsch.getSession(this.sshUsername, hostTracker.getResolvedAddress().getHostName(), this.sshPort);
        if (this.sshPassword == null) {
            jsch.addIdentity(this.sshUsername, this.sshPrivateKey, null, null);
        } else {
            session.setPassword(this.sshPassword);
        }
        session.setConfig(this.sshOptions);
        session.connect(shorterTimeout);
    } catch (JSchException e) {
        multipleDeclareDeployingNodeLost(depNodeURLs, "unable to " + msg + "\n" + getStackTraceAsString(e));
        throw new RMException("unable to " + msg, e);
    }
    SSHInfrastructureV2.logger.info("Executing SSH command: '" + finalCmdLine + "'");
    ScheduledExecutorService deployService = Executors.newSingleThreadScheduledExecutor();
    try {
        // Create ssh channel to run the cmd
        ByteArrayOutputStream baos = new ByteArrayOutputStream(DEFAULT_OUTPUT_BUFFER_LENGTH);
        ChannelExec channel;
        try {
            channel = (ChannelExec) session.openChannel("exec");
            channel.setCommand(finalCmdLine);
            channel.setOutputStream(baos);
            channel.setErrStream(baos);
            channel.connect();
        } catch (JSchException e) {
            multipleDeclareDeployingNodeLost(depNodeURLs, "unable to " + msg + "\n" + getStackTraceAsString(e));
            throw new RMException("unable to " + msg, e);
        }
        final ChannelExec chan = channel;
        Future<Void> deployResult = deployService.submit(new Callable<Void>() {

            @Override
            public Void call() throws Exception {
                while (!shutDown.get() && !checkAllNodesAreAcquiredAndDo(createdNodeNames, null, null)) {
                    if (anyTimedOut(depNodeURLs)) {
                        throw new IllegalStateException("The upper infrastructure has issued a timeout");
                    }
                    // processes live completely independently
                    if (!deployNodesInDetachedMode && chan.getExitStatus() != PROCESS_STILL_RUNNING_VALUE) {
                        throw new IllegalStateException("The jvm process of the node has exited prematurely");
                    }
                    try {
                        Thread.sleep(1000);
                    } catch (InterruptedException e) {
                        // we know the cause of this
                        return null;
                    // interruption just exit
                    }
                }
                // Victory
                return null;
            }
        });
        try {
            deployResult.get(shorterTimeout, TimeUnit.MILLISECONDS);
        } catch (ExecutionException e) {
            declareLostAndThrow("Unable to " + msg + " due to " + e.getCause(), depNodeURLs, channel, baos, e);
        } catch (InterruptedException e) {
            deployResult.cancel(true);
            declareLostAndThrow("Unable to " + msg + " due to an interruption", depNodeURLs, channel, baos, e);
        } catch (TimeoutException e) {
            deployResult.cancel(true);
            declareLostAndThrow("Unable to " + msg + " due to timeout", depNodeURLs, channel, baos, e);
        } finally {
            channel.disconnect();
        }
    } finally {
        removeTimeouts(depNodeURLs);
        session.disconnect();
        deployService.shutdownNow();
    }
}
Also used : JSchException(com.jcraft.jsch.JSchException) Throwables.getStackTraceAsString(com.google.common.base.Throwables.getStackTraceAsString) CommandLineBuilder(org.ow2.proactive.resourcemanager.utils.CommandLineBuilder) JSch(com.jcraft.jsch.JSch) RMException(org.ow2.proactive.resourcemanager.exception.RMException) Client(org.ow2.proactive.resourcemanager.authentication.Client) ExecutionException(java.util.concurrent.ExecutionException) TimeoutException(java.util.concurrent.TimeoutException) ScheduledExecutorService(java.util.concurrent.ScheduledExecutorService) IOException(java.io.IOException) ByteArrayOutputStream(java.io.ByteArrayOutputStream) KeyException(java.security.KeyException) ChannelExec(com.jcraft.jsch.ChannelExec) KeyException(java.security.KeyException) TimeoutException(java.util.concurrent.TimeoutException) IOException(java.io.IOException) RMException(org.ow2.proactive.resourcemanager.exception.RMException) ExecutionException(java.util.concurrent.ExecutionException) JSchException(com.jcraft.jsch.JSchException) Session(com.jcraft.jsch.Session)

Aggregations

Test (org.junit.Test)69 ISchedulerClient (org.ow2.proactive.scheduler.rest.ISchedulerClient)36 Client (org.ow2.proactive.resourcemanager.authentication.Client)31 JobId (org.ow2.proactive.scheduler.common.job.JobId)30 TaskFlowJob (org.ow2.proactive.scheduler.common.job.TaskFlowJob)28 NonTerminatingJob (functionaltests.jobs.NonTerminatingJob)25 SimpleJob (functionaltests.jobs.SimpleJob)25 Job (org.ow2.proactive.scheduler.common.job.Job)25 ResteasyClient (org.jboss.resteasy.client.jaxrs.ResteasyClient)18 ResteasyWebTarget (org.jboss.resteasy.client.jaxrs.ResteasyWebTarget)18 ListFile (org.ow2.proactive_grid_cloud_portal.dataspace.dto.ListFile)18 File (java.io.File)17 RMNode (org.ow2.proactive.resourcemanager.rmnode.RMNode)13 TaskResult (org.ow2.proactive.scheduler.common.task.TaskResult)13 NotConnectedRestException (org.ow2.proactive_grid_cloud_portal.scheduler.exception.NotConnectedRestException)11 Node (org.objectweb.proactive.core.node.Node)9 IOException (java.io.IOException)8 NotConnectedException (org.ow2.proactive.scheduler.common.exception.NotConnectedException)8 JobInfo (org.ow2.proactive.scheduler.common.job.JobInfo)8 NodeSet (org.ow2.proactive.utils.NodeSet)7