Slave.java 19.5 KB
Newer Older
K
kohsuke 已提交
1 2 3 4
package hudson.model;

import hudson.FilePath;
import hudson.Launcher;
5
import hudson.Launcher.RemoteLauncher;
K
kohsuke 已提交
6
import hudson.Util;
7 8
import hudson.tasks.DynamicLabeler;
import hudson.tasks.LabelFinder;
9 10
import hudson.maven.agent.Main;
import hudson.maven.agent.PluginManagerInterceptor;
K
kohsuke 已提交
11
import hudson.model.Descriptor.FormException;
K
kohsuke 已提交
12 13
import hudson.remoting.Callable;
import hudson.remoting.Channel;
14
import hudson.remoting.Channel.Listener;
K
kohsuke 已提交
15
import hudson.remoting.VirtualChannel;
16
import hudson.remoting.Which;
17
import hudson.util.NullStream;
18
import hudson.util.RingBufferLogHandler;
K
kohsuke 已提交
19 20
import hudson.util.StreamCopyThread;
import hudson.util.StreamTaskListener;
21 22
import org.kohsuke.stapler.StaplerRequest;
import org.kohsuke.stapler.StaplerResponse;
K
kohsuke 已提交
23

24 25
import javax.servlet.ServletException;
import javax.servlet.http.HttpServletResponse;
K
kohsuke 已提交
26
import java.io.File;
27 28
import java.io.FileNotFoundException;
import java.io.FileOutputStream;
K
kohsuke 已提交
29 30 31
import java.io.IOException;
import java.io.InputStream;
import java.io.OutputStream;
32
import java.io.PrintWriter;
33
import java.io.Serializable;
34 35
import java.net.URL;
import java.net.URLConnection;
36 37 38 39 40
import java.util.ArrayList;
import java.util.Collections;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
K
kohsuke 已提交
41
import java.util.logging.Level;
K
kohsuke 已提交
42
import java.util.logging.LogRecord;
43
import java.util.logging.Logger;
K
kohsuke 已提交
44

K
kohsuke 已提交
45 46 47 48 49
/**
 * Information about a Hudson slave node.
 *
 * @author Kohsuke Kawaguchi
 */
K
kohsuke 已提交
50
public final class Slave implements Node, Serializable {
K
kohsuke 已提交
51
    /**
K
kohsuke 已提交
52
     * Name of this slave node.
K
kohsuke 已提交
53
     */
K
kohsuke 已提交
54
    protected final String name;
K
kohsuke 已提交
55 56 57 58 59 60 61 62

    /**
     * Description of this node.
     */
    private final String description;

    /**
     * Path to the root of the workspace
K
kohsuke 已提交
63
     * from the view point of this node, such as "/hudson"
K
kohsuke 已提交
64
     */
K
kohsuke 已提交
65
    protected final String remoteFS;
K
kohsuke 已提交
66 67 68 69 70 71 72 73 74 75 76

    /**
     * Number of executors of this node.
     */
    private int numExecutors = 2;

    /**
     * Job allocation strategy.
     */
    private Mode mode;

K
kohsuke 已提交
77 78 79 80 81 82
    /**
     * Command line to launch the agent, like
     * "ssh myslave java -jar /path/to/hudson-remoting.jar"
     */
    private String agentCommand;

83 84 85 86 87 88 89 90 91 92
    /**
     * Whitespace-separated labels.
     */
    private String label="";

    /**
     * Lazily computed set of labels from {@link #label}.
     */
    private transient volatile Set<Label> labels;

93 94 95
    private transient volatile Set<Label> dynamicLabels;
    private transient volatile int dynamicLabelsInstanceHash;

K
kohsuke 已提交
96 97 98
    /**
     * @stapler-constructor
     */
99
    public Slave(String name, String description, String command, String remoteFS, int numExecutors, Mode mode,
100
                 String label) throws FormException {
K
kohsuke 已提交
101 102 103 104
        this.name = name;
        this.description = description;
        this.numExecutors = numExecutors;
        this.mode = mode;
K
kohsuke 已提交
105 106
        this.agentCommand = command;
        this.remoteFS = remoteFS;
107 108
        this.label = Util.fixNull(label).trim();
        getAssignedLabels();    // compute labels now
K
kohsuke 已提交
109

K
d'oh!  
kohsuke 已提交
110
        if (name.equals(""))
K
kohsuke 已提交
111
            throw new FormException("Invalid slave configuration. Name is empty", null);
K
kohsuke 已提交
112

113 114 115 116 117
        // this prevents the config from being saved when slaves are offline.
        // on a large deployment with a lot of slaves, some slaves are bound to be offline,
        // so this check is harmful.
        //if (!localFS.exists())
        //    throw new FormException("Invalid slave configuration for " + name + ". No such directory exists: " + localFS, null);
K
kohsuke 已提交
118 119
        if (remoteFS.equals(""))
            throw new FormException("Invalid slave configuration for " + name + ". No remote directory given", null);
K
kohsuke 已提交
120 121 122
    }

    public String getCommand() {
K
kohsuke 已提交
123
        return agentCommand;
K
kohsuke 已提交
124 125 126 127 128 129
    }

    public String getRemoteFS() {
        return remoteFS;
    }

K
kohsuke 已提交
130 131
    public String getNodeName() {
        return name;
K
kohsuke 已提交
132 133 134 135 136 137
    }

    public String getNodeDescription() {
        return description;
    }

K
kohsuke 已提交
138
    /**
139
     * Gets the root directory of the Hudson workspace on this slave.
K
kohsuke 已提交
140
     */
K
kohsuke 已提交
141
    public FilePath getFilePath() {
K
kohsuke 已提交
142
        return new FilePath(getComputer().getChannel(),remoteFS);
K
kohsuke 已提交
143 144 145 146 147 148 149 150 151 152
    }

    public int getNumExecutors() {
        return numExecutors;
    }

    public Mode getMode() {
        return mode;
    }

153 154 155
    public String getLabelString() {
        return Util.fixNull(label).trim();
    }
156

157
    public Set<Label> getAssignedLabels() {
158 159
        // todo refactor to make dynamic labels a bit less hacky
        if(labels==null || isChangedDynamicLabels()) {
160 161 162 163 164 165 166 167
            Set<Label> r = new HashSet<Label>();
            String ls = getLabelString();
            if(ls.length()>0) {
                for( String l : ls.split(" +")) {
                    r.add(Hudson.getInstance().getLabel(l));
                }
            }
            r.add(getSelfLabel());
168
            r.addAll(getDynamicLabels());
169 170 171 172 173
            this.labels = Collections.unmodifiableSet(r);
        }
        return labels;
    }

174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219
    /**
     * Check if we should rebuild the list of dynamic labels.
     * @todo make less hacky
     * @return
     */
    private boolean isChangedDynamicLabels() {
        Computer comp = getComputer();
        if (comp == null) {
            return dynamicLabelsInstanceHash != 0;
        } else {
            if (dynamicLabelsInstanceHash == comp.hashCode()) {
                return false;
            }
            dynamicLabels = null; // force a re-calc
            return true;
        }
    }

    /**
     * Returns the possibly empty set of labels that it has been determined as supported by this node.
     *
     * @todo make less hacky
     * @see hudson.tasks.LabelFinder
     */
    public Set<Label> getDynamicLabels() {
        if (dynamicLabels == null) {
            synchronized (this) {
                if (dynamicLabels == null) {
                    dynamicLabels = new HashSet<Label>();
                    Computer computer = getComputer();
                    VirtualChannel channel;
                    if (computer != null && (channel = computer.getChannel()) != null) {
                        dynamicLabelsInstanceHash = computer.hashCode();
                        for (DynamicLabeler labeler : LabelFinder.LABELERS) {
                            for (String label : labeler.findLabels(channel)) {
                                dynamicLabels.add(Hudson.getInstance().getLabel(label));
                            }
                        }
                    } else {
                        dynamicLabelsInstanceHash = 0;
                    }
                }
            }
        }
        return dynamicLabels;
    }
220 221 222 223 224

    public Label getSelfLabel() {
        return Hudson.getInstance().getLabel(name);
    }

K
kohsuke 已提交
225
    public long getClockDifference() throws IOException, InterruptedException {
K
kohsuke 已提交
226 227
        VirtualChannel channel = getComputer().getChannel();
        if(channel==null)   return 0;   // can't check
K
kohsuke 已提交
228

K
kohsuke 已提交
229 230 231 232 233 234 235
        long startTime = System.currentTimeMillis();
        long slaveTime = channel.call(new Callable<Long,RuntimeException>() {
            public Long call() {
                return System.currentTimeMillis();
            }
        });
        long endTime = System.currentTimeMillis();
K
kohsuke 已提交
236

K
kohsuke 已提交
237
        return (startTime+endTime)/2 - slaveTime;
K
kohsuke 已提交
238 239
    }

K
kohsuke 已提交
240

K
kohsuke 已提交
241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263
    /**
     * Gets the clock difference in HTML string.
     */
    public String getClockDifferenceString() {
        try {
            long diff = getClockDifference();
            if(-1000<diff && diff <1000)
                return "In sync";  // clock is in sync

            long abs = Math.abs(diff);

            String s = Util.getTimeSpanString(abs);
            if(diff<0)
                s += " ahead";
            else
                s += " behind";

            if(abs>100*60) // more than a minute difference
                s = "<span class='error'>"+s+"</span>";

            return s;
        } catch (IOException e) {
            return "<span class='error'>Unable to check</span>";
K
kohsuke 已提交
264 265
        } catch (InterruptedException e) {
            return "<span class='error'>Unable to check</span>";
K
kohsuke 已提交
266 267 268
        }
    }

K
kohsuke 已提交
269 270 271 272
    public Computer createComputer() {
        return new ComputerImpl(this);
    }

273 274 275 276
    public FilePath getWorkspaceFor(TopLevelItem item) {
        return getWorkspaceRoot().child(item.getName());
    }

K
kohsuke 已提交
277 278 279 280 281 282
    /**
     * Root directory on this slave where all the job workspaces are laid out.
     */
    public FilePath getWorkspaceRoot() {
        return getFilePath().child("workspace");
    }
K
kohsuke 已提交
283

K
kohsuke 已提交
284 285
    public static final class ComputerImpl extends Computer {
        private volatile Channel channel;
K
kohsuke 已提交
286

K
kohsuke 已提交
287 288 289 290 291 292
        /**
         * This is where the log from the remote agent goes.
         */
        private File getLogFile() {
            return new File(Hudson.getInstance().getRootDir(),"slave-"+nodeName+".log");
        }
K
kohsuke 已提交
293

K
kohsuke 已提交
294 295 296 297
        private ComputerImpl(Slave slave) {
            super(slave);
        }

298 299 300 301 302 303 304 305 306
        public Slave getNode() {
            return (Slave)super.getNode();
        }

        @Override
        public boolean isJnlpAgent() {
            return getNode().getCommand().length()==0;
        }

K
kohsuke 已提交
307 308 309 310 311 312
        /**
         * Launches a remote agent.
         */
        private void launch(final Slave slave) {
            closeChannel();

K
kohsuke 已提交
313
            final OutputStream launchLog = openLogFile();
K
kohsuke 已提交
314

315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341
            if(slave.agentCommand.length()>0) {
                // launch the slave agent asynchronously
                threadPoolForRemoting.execute(new Runnable() {
                    // TODO: do this only for nodes that are so configured.
                    // TODO: support passive connection via JNLP
                    public void run() {
                        final StreamTaskListener listener = new StreamTaskListener(launchLog);
                        try {
                            listener.getLogger().println("Launching slave agent");
                            listener.getLogger().println("$ "+slave.agentCommand);
                            final Process proc = Runtime.getRuntime().exec(slave.agentCommand);

                            // capture error information from stderr. this will terminate itself
                            // when the process is killed.
                            new StreamCopyThread("stderr copier for remote agent on "+slave.getNodeName(),
                                proc.getErrorStream(), launchLog).start();

                            setChannel(proc.getInputStream(),proc.getOutputStream(),launchLog,new Listener() {
                                public void onClosed(Channel channel, IOException cause) {
                                    if(cause!=null)
                                        cause.printStackTrace(listener.error("slave agent was terminated"));
                                    proc.destroy();
                                }
                            });

                            logger.info("slave agent launched for "+slave.getNodeName());

342 343
                        } catch (InterruptedException e) {
                            e.printStackTrace(listener.error("aborted"));
344 345 346 347 348 349 350 351 352 353
                        } catch (IOException e) {
                            Util.displayIOException(e,listener);

                            String msg = Util.getWin32ErrorMessage(e);
                            if(msg==null)   msg="";
                            else            msg=" : "+msg;
                            msg = "Unable to launch the slave agent for " + slave.getNodeName() + msg;
                            logger.log(Level.SEVERE,msg,e);
                            e.printStackTrace(listener.error(msg));
                        }
K
kohsuke 已提交
354
                    }
355 356
                });
            }
K
kohsuke 已提交
357
        }
K
kohsuke 已提交
358

K
kohsuke 已提交
359 360 361 362 363 364 365 366 367 368 369
        public OutputStream openLogFile() {
            OutputStream os;
            try {
                os = new FileOutputStream(getLogFile());
            } catch (FileNotFoundException e) {
                logger.log(Level.SEVERE, "Failed to create log file "+getLogFile(),e);
                os = new NullStream();
            }
            return os;
        }

370 371
        private final Object channelLock = new Object();

K
kohsuke 已提交
372 373 374
        /**
         * Creates a {@link Channel} from the given stream and sets that to this slave.
         */
375
        public void setChannel(InputStream in, OutputStream out, OutputStream launchLog, Listener listener) throws IOException, InterruptedException {
376
            synchronized(channelLock) {
K
kohsuke 已提交
377 378 379
                if(this.channel!=null)
                    throw new IllegalStateException("Already connected");

380
                Channel channel = new Channel(nodeName,threadPoolForRemoting,
K
kohsuke 已提交
381 382 383 384 385 386 387
                    in,out, launchLog);
                channel.addListener(new Listener() {
                    public void onClosed(Channel c,IOException cause) {
                        ComputerImpl.this.channel = null;
                    }
                });
                channel.addListener(listener);
388 389 390 391 392 393

                {// send jars that we need for our operations
                    // TODO: maybe I should generalize this kind of "post initialization" processing
                    PrintWriter log = new PrintWriter(launchLog,true);
                    FilePath dst = new FilePath(channel,getNode().getRemoteFS());
                    new FilePath(Which.jarFile(Main.class)).copyTo(dst.child("maven-agent.jar"));
394
                    log.println("Copied maven-agent.jar");
395
                    new FilePath(Which.jarFile(PluginManagerInterceptor.class)).copyTo(dst.child("maven-interceptor.jar"));
396
                    log.println("Copied maven-interceptor.jar");
397 398
                }

K
kohsuke 已提交
399
                // install log handler
K
kohsuke 已提交
400
                channel.call(new LogInstaller());
K
kohsuke 已提交
401 402


403 404
                // prevent others from seeing a channel that's not properly initialized yet
                this.channel = channel;
K
kohsuke 已提交
405 406 407 408
            }
            Hudson.getInstance().getQueue().scheduleMaintenance();
        }

K
kohsuke 已提交
409 410 411 412 413
        @Override
        public VirtualChannel getChannel() {
            return channel;
        }

K
kohsuke 已提交
414 415 416 417 418 419 420 421 422 423 424
        public List<LogRecord> getLogRecords() throws IOException, InterruptedException {
            if(channel==null)
                return Collections.emptyList();
            else
                return channel.call(new Callable<List<LogRecord>,RuntimeException>() {
                    public List<LogRecord> call() {
                        return new ArrayList<LogRecord>(SLAVE_LOG_HANDLER.getView());
                    }
                });
        }

K
kohsuke 已提交
425 426 427 428 429 430 431
        public void doDoDisconnect(StaplerRequest req, StaplerResponse rsp) throws IOException, ServletException {
            if(!Hudson.adminCheck(req,rsp))
                return;
            closeChannel();
            rsp.sendRedirect(".");
        }

K
kohsuke 已提交
432 433 434 435
        public void doLaunchSlaveAgent(StaplerRequest req, StaplerResponse rsp) throws IOException, ServletException {
            if(channel!=null) {
                rsp.sendError(HttpServletResponse.SC_NOT_FOUND);
                return;
K
kohsuke 已提交
436 437
            }

K
kohsuke 已提交
438
            launch(getNode());
K
kohsuke 已提交
439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458

            // TODO: would be nice to redirect the user to "launching..." wait page,
            // then spend a few seconds there and poll for the completion periodically.
            rsp.sendRedirect("log");
        }

        /**
         * Gets the string representation of the slave log.
         */
        public String getLog() throws IOException {
            return Util.loadFile(getLogFile());
        }

        /**
         * Handles incremental log.
         */
        public void doProgressiveLog( StaplerRequest req, StaplerResponse rsp) throws IOException {
            new LargeText(getLogFile(),false).doProgressText(req,rsp);
        }

459 460 461
        /**
         * Serves jar files for JNLP slave agents.
         */
K
kohsuke 已提交
462 463
        public JnlpJar getJnlpJars(String fileName) {
            return new JnlpJar(fileName);
464 465
        }

K
kohsuke 已提交
466 467 468 469 470 471 472 473 474 475 476 477 478 479
        @Override
        protected void kill() {
            super.kill();
            closeChannel();
        }

        private void closeChannel() {
            Channel c = channel;
            channel = null;
            if(c!=null)
                try {
                    c.close();
                } catch (IOException e) {
                    logger.log(Level.SEVERE, "Failed to terminate channel to "+getDisplayName(),e);
K
kohsuke 已提交
480
                }
K
kohsuke 已提交
481 482 483 484 485 486 487 488 489 490 491 492 493
        }

        @Override
        protected void setNode(Node node) {
            super.setNode(node);
            if(channel==null)
                // maybe the configuration was changed to relaunch the slave, so try it now.
                launch((Slave)node);
        }

        private static final Logger logger = Logger.getLogger(ComputerImpl.class.getName());
    }

494 495 496 497
    /**
     * Web-bound object used to serve jar files for JNLP.
     */
    public static final class JnlpJar {
K
kohsuke 已提交
498
        private final String fileName;
499

K
kohsuke 已提交
500 501
        public JnlpJar(String fileName) {
            this.fileName = fileName;
502 503 504
        }

        public void doIndex( StaplerRequest req, StaplerResponse rsp) throws IOException, ServletException {
K
kohsuke 已提交
505 506 507 508
            URL res = req.getServletContext().getResource("/WEB-INF/" + fileName);
            if(res==null) {
                // during the development this path doesn't have the files.
                res = new URL(new File(".").getAbsoluteFile().toURL(),"target/generated-resources/WEB-INF/"+fileName);
509 510
            }

K
kohsuke 已提交
511
            URLConnection con = res.openConnection();
512 513 514 515 516 517 518
            InputStream in = con.getInputStream();
            rsp.serveFile(req, in, con.getLastModified(), con.getContentLength(), "*.jar" );
            in.close();
        }

    }

K
kohsuke 已提交
519
    public Launcher createLauncher(TaskListener listener) {
520
        // Windows absolute path always include ':', but this is not a valid char in Unix file systems.
521 522
        // Windows can handle '/' as a path separator but Unix can't,
        // so err on Unix side
523
        boolean isUnix = !remoteFS.contains(":") && !remoteFS.contains("\\");
524

525
        return new RemoteLauncher(listener, getComputer().getChannel(),isUnix);
K
kohsuke 已提交
526 527
    }

K
kohsuke 已提交
528 529 530 531 532
    /**
     * Gets th ecorresponding computer object.
     */
    public Computer getComputer() {
        return Hudson.getInstance().getComputer(getNodeName());
K
kohsuke 已提交
533 534 535 536 537 538 539 540 541 542 543 544 545 546 547
    }

    public boolean equals(Object o) {
        if (this == o) return true;
        if (o == null || getClass() != o.getClass()) return false;

        final Slave that = (Slave) o;

        return name.equals(that.name);
    }

    public int hashCode() {
        return name.hashCode();
    }

K
kohsuke 已提交
548 549 550 551 552 553 554 555 556 557 558 559
    /**
     * Invoked by XStream when this object is read into memory.
     */
    private Object readResolve() {
        // convert the old format to the new one
        if(command!=null && agentCommand==null) {
            if(command.length()>0)  command += ' ';
            agentCommand = command+"java -jar ~/bin/slave.jar";
        }
        return this;
    }

K
kohsuke 已提交
560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
    /**
     * This field is used on each slave node to record log records on the slave.
     */
    private static final RingBufferLogHandler SLAVE_LOG_HANDLER = new RingBufferLogHandler();

    private static class LogInstaller implements Callable<Void,RuntimeException> {
        public Void call() {
            // avoid double installation of the handler
            Logger logger = Logger.getLogger("hudson");
            logger.removeHandler(SLAVE_LOG_HANDLER);
            logger.addHandler(SLAVE_LOG_HANDLER);
            return null;
        }
        private static final long serialVersionUID = 1L;
    }

K
kohsuke 已提交
576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594
//
// backwrad compatibility
//
    /**
     * In Hudson < 1.69 this was used to store the local file path
     * to the remote workspace. No longer in use.
     *
     * @deprecated
     *      ... but still in use during the transition.
     */
    private File localFS;

    /**
     * In Hudson < 1.69 this was used to store the command
     * to connect to the remote machine, like "ssh myslave".
     *
     * @deprecated
     */
    private transient String command;
K
kohsuke 已提交
595
}