private List<String> buildPySparkShellCommand(Map<String, String> env) throws IOException {
// For backwards compatibility, if a script is specified in
// the pyspark command line, then run it using spark-submit.
if (!appArgs.isEmpty() && appArgs.get(0).endsWith(".py")) {
System.err.println(
"Running python applications through 'pyspark' is not supported as of Spark 2.0.\n"
+ "Use ./bin/spark-submit <python file>");
System.exit(-1);
}
checkArgument(appArgs.isEmpty(), "pyspark does not support any application options.");
// When launching the pyspark shell, the spark-submit arguments should be stored in the
// PYSPARK_SUBMIT_ARGS env variable.
constructEnvVarArgs(env, "PYSPARK_SUBMIT_ARGS");
// The executable is the PYSPARK_DRIVER_PYTHON env variable set by the pyspark script,
// followed by PYSPARK_DRIVER_PYTHON_OPTS.
List<String> pyargs = new ArrayList<>();
pyargs.add(firstNonEmpty(System.getenv("PYSPARK_DRIVER_PYTHON"), "python"));
String pyOpts = System.getenv("PYSPARK_DRIVER_PYTHON_OPTS");
if (!isEmpty(pyOpts)) {
pyargs.addAll(parseOptionString(pyOpts));
}
return pyargs;
}
/**
* Builds a list of arguments to run java.
*
* <p>This method finds the java executable to use and appends JVM-specific options for running a
* class with Spark in the classpath. It also loads options from the "java-opts" file in the
* configuration directory being used.
*
* <p>Callers should still add at least the class to run, as well as any arguments to pass to the
* class.
*/
List<String> buildJavaCommand(String extraClassPath) throws IOException {
List<String> cmd = new ArrayList<String>();
String envJavaHome;
if (javaHome != null) {
cmd.add(join(File.separator, javaHome, "bin", "java"));
} else if ((envJavaHome = System.getenv("JAVA_HOME")) != null) {
cmd.add(join(File.separator, envJavaHome, "bin", "java"));
} else {
cmd.add(join(File.separator, System.getProperty("java.home"), "bin", "java"));
}
// Load extra JAVA_OPTS from conf/java-opts, if it exists.
File javaOpts = new File(join(File.separator, getConfDir(), "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();
}
}
cmd.add("-cp");
cmd.add(join(File.pathSeparator, buildClassPath(extraClassPath)));
return cmd;
}
private List<String> buildSparkRCommand(Map<String, String> env) throws IOException {
if (!appArgs.isEmpty() && appArgs.get(0).endsWith(".R")) {
System.err.println(
"Running R applications through 'sparkR' is not supported as of Spark 2.0.\n"
+ "Use ./bin/spark-submit <R file>");
System.exit(-1);
}
// When launching the SparkR shell, store the spark-submit arguments in the SPARKR_SUBMIT_ARGS
// env variable.
constructEnvVarArgs(env, "SPARKR_SUBMIT_ARGS");
// Set shell.R as R_PROFILE_USER to load the SparkR package when the shell comes up.
String sparkHome = System.getenv("SPARK_HOME");
env.put(
"R_PROFILE_USER",
join(File.separator, sparkHome, "R", "lib", "SparkR", "profile", "shell.R"));
List<String> args = new ArrayList<>();
args.add(firstNonEmpty(System.getenv("SPARKR_DRIVER_R"), "R"));
return args;
}
private List<String> buildSparkSubmitCommand(Map<String, String> env) throws IOException {
// Load the properties file and check whether spark-submit will be running the app's driver
// or just launching a cluster app. When running the driver, the JVM's argument will be
// modified to cover the driver's configuration.
Map<String, String> config = getEffectiveConfig();
boolean isClientMode = isClientMode(config);
String extraClassPath = isClientMode ? config.get(SparkLauncher.DRIVER_EXTRA_CLASSPATH) : null;
List<String> cmd = buildJavaCommand(extraClassPath);
// Take Thrift Server as daemon
if (isThriftServer(mainClass)) {
addOptionString(cmd, System.getenv("SPARK_DAEMON_JAVA_OPTS"));
}
addOptionString(cmd, System.getenv("SPARK_SUBMIT_OPTS"));
addOptionString(cmd, System.getenv("SPARK_JAVA_OPTS"));
if (isClientMode) {
// Figuring out where the memory value come from is a little tricky due to precedence.
// Precedence is observed in the following order:
// - explicit configuration (setConf()), which also covers --driver-memory cli argument.
// - properties file.
// - SPARK_DRIVER_MEMORY env variable
// - SPARK_MEM env variable
// - default value (1g)
// Take Thrift Server as daemon
String tsMemory = isThriftServer(mainClass) ? System.getenv("SPARK_DAEMON_MEMORY") : null;
String memory =
firstNonEmpty(
tsMemory,
config.get(SparkLauncher.DRIVER_MEMORY),
System.getenv("SPARK_DRIVER_MEMORY"),
System.getenv("SPARK_MEM"),
DEFAULT_MEM);
cmd.add("-Xms" + memory);
cmd.add("-Xmx" + memory);
addOptionString(cmd, config.get(SparkLauncher.DRIVER_EXTRA_JAVA_OPTIONS));
mergeEnvPathList(
env, getLibPathEnvName(), config.get(SparkLauncher.DRIVER_EXTRA_LIBRARY_PATH));
}
addPermGenSizeOpt(cmd);
cmd.add("org.apache.spark.deploy.SparkSubmit");
cmd.addAll(buildSparkSubmitArgs());
return cmd;
}
String getenv(String key) {
return firstNonEmpty(childEnv.get(key), System.getenv(key));
}