Slave.java 13.4 KB
Newer Older
K
kohsuke 已提交
1 2
package hudson.model;

3
import hudson.CloseProofOutputStream;
K
kohsuke 已提交
4 5
import hudson.FilePath;
import hudson.Launcher;
6
import hudson.Launcher.LocalLauncher;
K
kohsuke 已提交
7
import hudson.Proc;
K
kohsuke 已提交
8
import hudson.Proc.RemoteProc;
K
kohsuke 已提交
9
import hudson.Util;
K
kohsuke 已提交
10
import hudson.model.Descriptor.FormException;
K
kohsuke 已提交
11 12
import hudson.remoting.Callable;
import hudson.remoting.Channel;
13
import hudson.remoting.Channel.Listener;
K
kohsuke 已提交
14 15 16
import hudson.remoting.RemoteInputStream;
import hudson.remoting.RemoteOutputStream;
import hudson.remoting.VirtualChannel;
17
import hudson.util.NullStream;
K
kohsuke 已提交
18 19
import hudson.util.StreamCopyThread;
import hudson.util.StreamTaskListener;
20 21
import org.kohsuke.stapler.StaplerRequest;
import org.kohsuke.stapler.StaplerResponse;
K
kohsuke 已提交
22

23 24
import javax.servlet.ServletException;
import javax.servlet.http.HttpServletResponse;
K
kohsuke 已提交
25
import java.io.File;
26 27
import java.io.FileNotFoundException;
import java.io.FileOutputStream;
K
kohsuke 已提交
28 29 30
import java.io.IOException;
import java.io.InputStream;
import java.io.OutputStream;
K
kohsuke 已提交
31 32 33 34
import java.io.Serializable;
import java.util.logging.Level;
import java.util.logging.Logger;

K
kohsuke 已提交
35 36 37 38 39
/**
 * Information about a Hudson slave node.
 *
 * @author Kohsuke Kawaguchi
 */
K
kohsuke 已提交
40
public final class Slave implements Node, Serializable {
K
kohsuke 已提交
41
    /**
42
     * PluginName of this slave node.
K
kohsuke 已提交
43
     */
K
kohsuke 已提交
44
    protected final String name;
K
kohsuke 已提交
45 46 47 48 49 50 51 52

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

    /**
     * Path to the root of the workspace
K
kohsuke 已提交
53
     * from the view point of this node, such as "/hudson"
K
kohsuke 已提交
54
     */
K
kohsuke 已提交
55
    protected final String remoteFS;
K
kohsuke 已提交
56 57 58 59 60 61 62 63 64 65 66

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

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

K
kohsuke 已提交
67 68 69 70 71 72 73 74 75 76
    /**
     * Command line to launch the agent, like
     * "ssh myslave java -jar /path/to/hudson-remoting.jar"
     */
    private String agentCommand;

    /**
     * @stapler-constructor
     */
    public Slave(String name, String description, String command, String remoteFS, int numExecutors, Mode mode) throws FormException {
K
kohsuke 已提交
77 78 79 80
        this.name = name;
        this.description = description;
        this.numExecutors = numExecutors;
        this.mode = mode;
K
kohsuke 已提交
81 82
        this.agentCommand = command;
        this.remoteFS = remoteFS;
K
kohsuke 已提交
83

K
d'oh!  
kohsuke 已提交
84
        if (name.equals(""))
85
            throw new FormException("Invalid slave configuration. PluginName is empty", null);
K
kohsuke 已提交
86

87 88 89 90 91
        // 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
92 93
        if (remoteFS.equals(""))
            throw new FormException("Invalid slave configuration for " + name + ". No remote directory given", null);
K
kohsuke 已提交
94 95 96
    }

    public String getCommand() {
K
kohsuke 已提交
97
        return agentCommand;
K
kohsuke 已提交
98 99 100 101 102 103
    }

    public String getRemoteFS() {
        return remoteFS;
    }

K
kohsuke 已提交
104 105
    public String getNodeName() {
        return name;
K
kohsuke 已提交
106 107 108 109 110 111 112
    }

    public String getNodeDescription() {
        return description;
    }

    public FilePath getFilePath() {
K
kohsuke 已提交
113
        return new FilePath(getComputer().getChannel(),remoteFS);
K
kohsuke 已提交
114 115 116 117 118 119 120 121 122 123 124 125 126 127 128
    }

    public int getNumExecutors() {
        return numExecutors;
    }

    public Mode getMode() {
        return mode;
    }

    /**
     * Estimates the clock difference with this slave.
     *
     * @return
     *      difference in milli-seconds.
K
kohsuke 已提交
129
     *      a positive value indicates that the master is ahead of the slave,
K
kohsuke 已提交
130 131 132
     *      and negative value indicates otherwise.
     */
    public long getClockDifference() throws IOException {
K
kohsuke 已提交
133 134
        VirtualChannel channel = getComputer().getChannel();
        if(channel==null)   return 0;   // can't check
K
kohsuke 已提交
135

K
kohsuke 已提交
136 137 138 139 140 141 142 143
        try {
            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
144

K
kohsuke 已提交
145 146 147 148
            return (startTime+endTime)/2 - slaveTime;
        } catch (InterruptedException e) {
            return 0;   // couldn't check
        }
K
kohsuke 已提交
149 150
    }

K
kohsuke 已提交
151

K
kohsuke 已提交
152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
    /**
     * 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
178 179 180 181
    public Computer createComputer() {
        return new ComputerImpl(this);
    }

182 183 184 185
    public FilePath getWorkspaceFor(TopLevelItem item) {
        return getWorkspaceRoot().child(item.getName());
    }

K
kohsuke 已提交
186 187 188 189 190 191
    /**
     * 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
192

K
kohsuke 已提交
193 194
    public static final class ComputerImpl extends Computer {
        private volatile Channel channel;
K
kohsuke 已提交
195

K
kohsuke 已提交
196 197 198 199 200 201
        /**
         * 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
202

K
kohsuke 已提交
203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218
        private ComputerImpl(Slave slave) {
            super(slave);
        }

        /**
         * Launches a remote agent.
         */
        private void launch(final Slave slave) {
            closeChannel();

            OutputStream os;
            try {
                os = new FileOutputStream(getLogFile());
            } catch (FileNotFoundException e) {
                logger.log(Level.SEVERE, "Failed to create log file "+getLogFile(),e);
                os = new NullStream();
K
kohsuke 已提交
219
            }
K
kohsuke 已提交
220 221 222 223 224 225 226 227 228 229 230
            final OutputStream launchLog = os;

            // 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);
K
kohsuke 已提交
231
                        final Process proc = Runtime.getRuntime().exec(slave.agentCommand);
K
kohsuke 已提交
232 233 234 235 236 237 238 239 240 241

                        // 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();

                        channel = new Channel(nodeName,threadPoolForRemoting,
                            proc.getInputStream(),proc.getOutputStream(), launchLog);
                        channel.addListener(new Listener() {
                            public void onClosed(Channel c,IOException cause) {
K
kohsuke 已提交
242 243
                                if(cause!=null)
                                    cause.printStackTrace(listener.error("slave agent was terminated"));
K
kohsuke 已提交
244
                                channel = null;
K
kohsuke 已提交
245
                                proc.destroy();
K
kohsuke 已提交
246 247 248 249
                            }
                        });

                        logger.info("slave agent launched for "+slave.getNodeName());
250 251

                        Hudson.getInstance().getQueue().scheduleMaintenance();
K
kohsuke 已提交
252 253 254 255 256 257 258 259 260 261 262 263 264
                    } 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
265

K
kohsuke 已提交
266 267 268 269 270 271 272 273 274
        @Override
        public VirtualChannel getChannel() {
            return channel;
        }

        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
275 276
            }

K
kohsuke 已提交
277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311
            launch((Slave) getNode());

            // 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);
        }

        @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
312
                }
K
kohsuke 已提交
313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342
        }

        @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());
    }

    public Launcher createLauncher(TaskListener listener) {
        return new Launcher(listener, getComputer().getChannel()) {
            public Proc launch(final String[] cmd, final String[] env, InputStream _in, OutputStream _out, FilePath _workDir) throws IOException {
                printCommandLine(cmd,_workDir);

                final OutputStream out = new RemoteOutputStream(new CloseProofOutputStream(_out));
                final InputStream  in  = _in==null ? null : new RemoteInputStream(_in);
                final String workDir = _workDir==null ? null : _workDir.getRemote();

                return new RemoteProc(getChannel().callAsync(new RemoteLaunchCallable(cmd, env, in, out, workDir)));
            }

            @Override
            public boolean isUnix() {
                // Windows can handle '/' as a path separator but Unix can't,
                // so err on Unix side
                return remoteFS.indexOf("\\")==-1;
K
kohsuke 已提交
343 344 345 346
            }
        };
    }

K
kohsuke 已提交
347 348 349 350 351
    /**
     * Gets th ecorresponding computer object.
     */
    public Computer getComputer() {
        return Hudson.getInstance().getComputer(getNodeName());
K
kohsuke 已提交
352 353 354 355 356 357 358 359 360 361 362 363 364 365 366
    }

    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
367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421
    /**
     * 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;
    }

//
// 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;

    private static class RemoteLaunchCallable implements Callable<Integer,IOException> {
        private final String[] cmd;
        private final String[] env;
        private final InputStream in;
        private final OutputStream out;
        private final String workDir;

        public RemoteLaunchCallable(String[] cmd, String[] env, InputStream in, OutputStream out, String workDir) {
            this.cmd = cmd;
            this.env = env;
            this.in = in;
            this.out = out;
            this.workDir = workDir;
        }

        public Integer call() throws IOException {
            Proc p = new LocalLauncher(TaskListener.NULL).launch(cmd, env, in, out,
                workDir ==null ? null : new FilePath(new File(workDir)));
            return p.join();
        }

        private static final long serialVersionUID = 1L;
    }
K
kohsuke 已提交
422
}