Github user andrewor14 commented on a diff in the pull request: https://github.com/apache/spark/pull/3916#discussion_r22841252 --- Diff: launcher/src/main/java/org/apache/spark/launcher/AbstractLauncher.java --- @@ -0,0 +1,600 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.launcher; + +import java.io.BufferedReader; +import java.io.File; +import java.io.FileFilter; +import java.io.FileInputStream; +import java.io.InputStreamReader; +import java.io.IOException; +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Enumeration; +import java.util.HashMap; +import java.util.List; +import java.util.Map; +import java.util.Properties; +import java.util.jar.JarEntry; +import java.util.jar.JarFile; +import java.util.regex.Pattern; + +/** + * Basic functionality for launchers. + */ +public abstract class AbstractLauncher<T extends AbstractLauncher> extends LauncherCommon { + + protected static final String DEFAULT_MEM = "512m"; + + protected String javaHome; + protected String sparkHome; + protected String propertiesFile; + protected final Map<String, String> conf = new HashMap<String, String>(); + private final Map<String, String> env; + + protected AbstractLauncher() { + this(null); + } + + protected AbstractLauncher(Map<String, String> env) { + this.env = env; + } + + @SuppressWarnings("unchecked") + private final T THIS = (T) this; + + /** Set a custom JAVA_HOME for launching the Spark application. */ + public T setJavaHome(String path) { + checkNotNull(path, "path"); + this.javaHome = path; + return THIS; + } + + /** Set a custom Spark installation location for the application. */ + public T setSparkHome(String path) { + checkNotNull(path, "path"); + this.sparkHome = path; + return THIS; + } + + /** Set a custom properties file with Spark configuration for the application. */ + public T setPropertiesFile(String path) { + checkNotNull(path, "path"); + this.propertiesFile = path; + return THIS; + } + + /** Set a single configuration value for the application. */ + public T setConf(String key, String value) { + checkNotNull(key, "key"); + checkNotNull(value, "value"); + checkArgument(key.startsWith("spark."), "'key' must start with 'spark.'"); + conf.put(key, value); + return THIS; + } + + /** + * Launchers should implement this to create the command to be executed. + */ + protected abstract List<String> buildLauncherCommand() throws IOException; + + protected Properties loadPropertiesFile() throws IOException { + Properties props = new Properties(); + File propsFile; + if (propertiesFile != null) { + propsFile = new File(propertiesFile); + checkArgument(propsFile.isFile(), "Invalid properties file '%s'.", propertiesFile); + } else { + String confDir = getenv("SPARK_CONF_DIR"); + if (confDir == null) { + confDir = join(File.separator, getSparkHome(), "conf"); + } + propsFile = new File(confDir, "spark-defaults.conf"); + } + + if (propsFile.isFile()) { + FileInputStream fd = null; + try { + fd = new FileInputStream(propsFile); + props.load(new InputStreamReader(fd, "UTF-8")); + } finally { + if (fd != null) { + try { + fd.close(); + } catch (IOException e) { + // Ignore. + } + } + } + } + + return props; + } + + protected String getSparkHome() { + String path = first(sparkHome, getenv("SPARK_HOME")); + checkState(path != null, + "Spark home not found; set it explicitly or use the SPARK_HOME environment variable."); + return path; + } + + protected List<String> createJavaCommand() throws IOException { + List<String> cmd = new ArrayList<String>(); + if (javaHome == null) { + cmd.add(join(File.separator, System.getProperty("java.home"), "..", "bin", "java")); + } else { + cmd.add(join(File.separator, javaHome, "bin", "java")); + } + + // Don't set MaxPermSize for Java 8 and later. + String[] version = System.getProperty("java.version").split("\\."); + if (Integer.parseInt(version[0]) == 1 && Integer.parseInt(version[1]) < 8) { + cmd.add("-XX:MaxPermSize=128m"); + } + + // Load extra JAVA_OPTS from conf/java-opts, if it exists. + File javaOpts = new File(join(File.separator, getSparkHome(), "conf", "java-opts")); + if (javaOpts.isFile()) { + BufferedReader br = new BufferedReader(new InputStreamReader( + new FileInputStream(javaOpts), "UTF-8")); + try { + String line; + while ((line = br.readLine()) != null) { + addOptionString(cmd, line); + } + } finally { + br.close(); + } + } + + return cmd; + } + + protected void addOptionString(List<String> cmd, String options) { + if (!isEmpty(options)) { + for (String opt : parseOptionString(options)) { + cmd.add(opt); + } + } + } + + /** + * Builds the classpath for the application. Returns a list with one classpath entry per element; + * each entry is formatted in the way expected by <i>java.net.URLClassLoader</i> (more + * specifically, with trailing slashes for directories). + */ + protected List<String> buildClassPath(String appClassPath) throws IOException { + String sparkHome = getSparkHome(); + String scala = getScalaVersion(); + + List<String> cp = new ArrayList<String>(); + addToClassPath(cp, getenv("SPARK_CLASSPATH")); + addToClassPath(cp, appClassPath); + + String confDir = getenv("SPARK_CONF_DIR"); + if (!isEmpty(confDir)) { + addToClassPath(cp, confDir); + } else { + addToClassPath(cp, join(File.separator, getSparkHome(), "conf")); + } + + boolean prependClasses = !isEmpty(getenv("SPARK_PREPEND_CLASSES")); + boolean isTesting = "1".equals(getenv("SPARK_TESTING")); + if (prependClasses || isTesting) { + List<String> projects = Arrays.asList("core", "repl", "mllib", "bagel", "graphx", + "streaming", "tools", "sql/catalyst", "sql/core", "sql/hive", "sql/hive-thriftserver", + "yarn", "launcher"); + if (prependClasses) { + System.err.println( + "NOTE: SPARK_PREPEND_CLASSES is set, placing locally compiled Spark classes ahead of " + + "assembly."); + for (String project : projects) { + addToClassPath(cp, String.format("%s/%s/target/scala-%s/classes", sparkHome, project, + scala)); + } + } + if (isTesting) { + for (String project : projects) { + addToClassPath(cp, String.format("%s/%s/target/scala-%s/test-classes", sparkHome, + project, scala)); + } + } + addToClassPath(cp, String.format("%s/core/target/jars/*", sparkHome)); + } + + String assembly = findAssembly(scala); + addToClassPath(cp, assembly); + + // When Hive support is needed, Datanucleus jars must be included on the classpath. Datanucleus + // jars do not work if only included in the uber jar as plugin.xml metadata is lost. Both sbt + // and maven will populate "lib_managed/jars/" with the datanucleus jars when Spark is built + // with Hive, so first check if the datanucleus jars exist, and then ensure the current Spark + // assembly is built for Hive, before actually populating the CLASSPATH with the jars. + // + // This block also serves as a check for SPARK-1703, when the assembly jar is built with + // Java 7 and ends up with too many files, causing issues with other JDK versions. + boolean needsDataNucleus = false; + JarFile assemblyJar = null; + try { + assemblyJar = new JarFile(assembly); + needsDataNucleus = assemblyJar.getEntry("org/apache/hadoop/hive/ql/exec/") != null; + } catch (IOException ioe) { + if (ioe.getMessage().indexOf("invalid CEN header") > 0) { + System.err.println( + "Loading Spark jar with '$JAR_CMD' failed.\n" + + "This is likely because Spark was compiled with Java 7 and run\n" + + "with Java 6 (see SPARK-1703). Please use Java 7 to run Spark\n" + + "or build Spark with Java 6."); + System.exit(1); + } else { + throw ioe; + } + } finally { + if (assemblyJar != null) { + try { + assemblyJar.close(); + } catch (IOException e) { + // Ignore. + } + } + } + + if (needsDataNucleus) { + System.err.println("Spark assembly has been built with Hive, including Datanucleus jars " + + "in classpath."); + File libdir; + if (new File(sparkHome, "RELEASE").isFile()) { + libdir = new File(sparkHome, "lib"); + } else { + libdir = new File(sparkHome, "lib_managed/jars"); + } + + checkState(libdir.isDirectory(), "Library directory '%s' does not exist.", + libdir.getAbsolutePath()); + for (File jar : libdir.listFiles()) { + if (jar.getName().startsWith("datanucleus-")) { + addToClassPath(cp, jar.getAbsolutePath()); + } + } + } + + addToClassPath(cp, getenv("HADOOP_CONF_DIR")); + addToClassPath(cp, getenv("YARN_CONF_DIR")); + addToClassPath(cp, getenv("SPARK_DIST_CLASSPATH")); + return cp; + } + + private void addToClassPath(List<String> cp, String entries) { --- End diff -- can you add java doc here to explain what `entries` look like, whether it mutates `cp` etc.
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org