HADOOP-16411. Fix javadoc warnings in hadoop-dynamometer.
Signed-off-by: Masatake Iwasaki <iwasakims@apache.org>
This commit is contained in:
parent
4632708148
commit
738c09349e
@ -90,19 +90,19 @@
|
||||
* part of this YARN application. This does not implement any retry/failure
|
||||
* handling.
|
||||
* TODO: Add proper retry/failure handling
|
||||
*
|
||||
* <p/>The AM will persist until it has run for a period of time equal to the
|
||||
* <p>
|
||||
* The AM will persist until it has run for a period of time equal to the
|
||||
* timeout specified or until the application is killed.
|
||||
*
|
||||
* <p/>If the NameNode is launched internally, it will upload some information
|
||||
* <p>
|
||||
* If the NameNode is launched internally, it will upload some information
|
||||
* onto the remote HDFS instance (i.e., the default FileSystem) about its
|
||||
* hostname and ports. This is in the location determined by the
|
||||
* {@link DynoConstants#DYNAMOMETER_STORAGE_DIR} and
|
||||
* {@link DynoConstants#NN_INFO_FILE_NAME} constants and is in the
|
||||
* {@link Properties} file format. This is consumed by this AM as well as the
|
||||
* {@link Client} to determine how to contact the NameNode.
|
||||
*
|
||||
* <p/>Information about the location of the DataNodes is logged by the AM.
|
||||
* <p>
|
||||
* Information about the location of the DataNodes is logged by the AM.
|
||||
*/
|
||||
@InterfaceAudience.Public
|
||||
@InterfaceStability.Unstable
|
||||
@ -204,6 +204,7 @@ public ApplicationMaster() {
|
||||
*
|
||||
* @param args Command line args
|
||||
* @return Whether init successful and run should be invoked
|
||||
* @throws ParseException on error while parsing options
|
||||
*/
|
||||
public boolean init(String[] args) throws ParseException {
|
||||
|
||||
@ -267,6 +268,9 @@ private void printUsage(Options opts) {
|
||||
*
|
||||
* @return True if the application completed successfully; false if if exited
|
||||
* unexpectedly, failed, was killed, etc.
|
||||
* @throws YarnException for issues while contacting YARN daemons
|
||||
* @throws IOException for other issues
|
||||
* @throws InterruptedException when the thread is interrupted
|
||||
*/
|
||||
public boolean run() throws YarnException, IOException, InterruptedException {
|
||||
LOG.info("Starting ApplicationMaster");
|
||||
|
@ -108,24 +108,24 @@
|
||||
* for them to be accessed by the YARN app, then launches an
|
||||
* {@link ApplicationMaster}, which is responsible for managing the lifetime of
|
||||
* the application.
|
||||
*
|
||||
* <p/>The Dynamometer YARN application starts up the DataNodes of an HDFS
|
||||
* <p>
|
||||
* The Dynamometer YARN application starts up the DataNodes of an HDFS
|
||||
* cluster. If the namenode_servicerpc_addr option is specified, it should point
|
||||
* to the service RPC address of an existing namenode, which the datanodes will
|
||||
* talk to. Else, a namenode will be launched internal to this YARN application.
|
||||
* The ApplicationMaster's logs contain links to the NN / DN containers to be
|
||||
* able to access their logs. Some of this information is also printed by the
|
||||
* client.
|
||||
*
|
||||
* <p/>The application will store files in the submitting user's home directory
|
||||
* <p>
|
||||
* The application will store files in the submitting user's home directory
|
||||
* under a `.dynamometer/applicationID/` folder. This is mostly for uses
|
||||
* internal to the application, but if the NameNode is launched through YARN,
|
||||
* the NameNode's metrics will also be uploaded to a file `namenode_metrics`
|
||||
* within this folder. This file is also accessible as part of the NameNode's
|
||||
* logs, but this centralized location is easier to access for subsequent
|
||||
* parsing.
|
||||
*
|
||||
* <p/>If the NameNode is launched internally, this Client will monitor the
|
||||
* <p>
|
||||
* If the NameNode is launched internally, this Client will monitor the
|
||||
* status of the NameNode, printing information about its availability as the
|
||||
* DataNodes register (e.g., outstanding under replicated blocks as block
|
||||
* reports arrive). If this is configured to launch the workload job, once the
|
||||
@ -134,8 +134,8 @@
|
||||
* NameNode. Once the workload job completes, the infrastructure application
|
||||
* will be shut down. At this time only the audit log replay
|
||||
* ({@link AuditReplayMapper}) workload is supported.
|
||||
*
|
||||
* <p/>If there is no workload job configured, this application will, by
|
||||
* <p>
|
||||
* If there is no workload job configured, this application will, by
|
||||
* default, persist indefinitely until killed by YARN. You can specify the
|
||||
* timeout option to have it exit automatically after some time. This timeout
|
||||
* will enforced if there is a workload job configured as well.
|
||||
@ -248,8 +248,8 @@ public class Client extends Configured implements Tool {
|
||||
private Options opts;
|
||||
|
||||
/**
|
||||
* @param args
|
||||
* Command line arguments
|
||||
* @param args Command line arguments
|
||||
* @throws Exception on error
|
||||
*/
|
||||
public static void main(String[] args) throws Exception {
|
||||
Client client = new Client(
|
||||
@ -386,6 +386,8 @@ private void printUsage() {
|
||||
*
|
||||
* @param args Parsed command line options
|
||||
* @return Whether the init was successful to run the client
|
||||
* @throws ParseException on error while parsing
|
||||
* @throws IOException for other errors
|
||||
*/
|
||||
public boolean init(String[] args) throws ParseException, IOException {
|
||||
|
||||
@ -506,6 +508,8 @@ public boolean init(String[] args) throws ParseException, IOException {
|
||||
* Main run function for the client.
|
||||
*
|
||||
* @return true if application completed successfully
|
||||
* @throws IOException for general issues
|
||||
* @throws YarnException for issues while contacting YARN daemons
|
||||
*/
|
||||
public boolean run() throws IOException, YarnException {
|
||||
|
||||
|
@ -116,9 +116,13 @@ private DynoInfraUtils() {}
|
||||
* (checked in that order) is set, use that as the mirror; else use
|
||||
* {@value APACHE_DOWNLOAD_MIRROR_DEFAULT}.
|
||||
*
|
||||
* @param destinationDir destination directory to save a tarball
|
||||
* @param version The version of Hadoop to download, like "2.7.4"
|
||||
* or "3.0.0-beta1"
|
||||
* @param conf configuration
|
||||
* @param log logger instance
|
||||
* @return The path to the tarball.
|
||||
* @throws IOException on failure
|
||||
*/
|
||||
public static File fetchHadoopTarball(File destinationDir, String version,
|
||||
Configuration conf, Logger log) throws IOException {
|
||||
|
@ -32,8 +32,8 @@
|
||||
/**
|
||||
* CreateFileMapper continuously creates 1 byte files for the specified duration
|
||||
* to increase the number of file objects on the NN.
|
||||
*
|
||||
* <p/>Configuration options available:
|
||||
* <p>
|
||||
* Configuration options available:
|
||||
* <ul>
|
||||
* <li>{@value NUM_MAPPERS_KEY} (required): Number of mappers to launch.</li>
|
||||
* <li>{@value DURATION_MIN_KEY} (required): Number of minutes to induce
|
||||
|
@ -34,6 +34,8 @@ public abstract class WorkloadMapper<KEYIN, VALUEIN>
|
||||
|
||||
/**
|
||||
* Return the input class to be used by this mapper.
|
||||
* @param conf configuration.
|
||||
* @return the {@link InputFormat} implementation for the mapper.
|
||||
*/
|
||||
public Class<? extends InputFormat> getInputFormat(Configuration conf) {
|
||||
return VirtualInputFormat.class;
|
||||
@ -41,18 +43,22 @@ public Class<? extends InputFormat> getInputFormat(Configuration conf) {
|
||||
|
||||
/**
|
||||
* Get the description of the behavior of this mapper.
|
||||
* @return description string.
|
||||
*/
|
||||
public abstract String getDescription();
|
||||
|
||||
/**
|
||||
* Get a list of the description of each configuration that this mapper
|
||||
* accepts.
|
||||
* @return list of the description of each configuration.
|
||||
*/
|
||||
public abstract List<String> getConfigDescriptions();
|
||||
|
||||
/**
|
||||
* Verify that the provided configuration contains all configurations required
|
||||
* by this mapper.
|
||||
* @param conf configuration.
|
||||
* @return whether or not all configurations required are provided.
|
||||
*/
|
||||
public abstract boolean verifyConfigurations(Configuration conf);
|
||||
|
||||
|
@ -35,6 +35,7 @@ public interface AuditCommandParser {
|
||||
* called prior to any calls to {@link #parse(Text, Function)}.
|
||||
*
|
||||
* @param conf The Configuration to be used to set up this parser.
|
||||
* @throws IOException if error on initializing a parser.
|
||||
*/
|
||||
void initialize(Configuration conf) throws IOException;
|
||||
|
||||
@ -50,6 +51,7 @@ public interface AuditCommandParser {
|
||||
* (in milliseconds) to absolute timestamps
|
||||
* (in milliseconds).
|
||||
* @return A command representing the input line.
|
||||
* @throws IOException if error on parsing.
|
||||
*/
|
||||
AuditReplayCommand parse(Text inputLine,
|
||||
Function<Long, Long> relativeToAbsolute) throws IOException;
|
||||
|
@ -36,8 +36,8 @@
|
||||
* It requires setting the {@value AUDIT_START_TIMESTAMP_KEY} configuration to
|
||||
* specify what the start time of the audit log was to determine when events
|
||||
* occurred relative to this start time.
|
||||
*
|
||||
* <p/>By default, this assumes that the audit log is in the default log format
|
||||
* <p>
|
||||
* By default, this assumes that the audit log is in the default log format
|
||||
* set up by Hadoop, like:
|
||||
* <pre>{@code
|
||||
* 1970-01-01 00:00:00,000 INFO FSNamesystem.audit: allowed=true ...
|
||||
|
@ -40,7 +40,9 @@
|
||||
* INSERT OVERWRITE DIRECTORY '${outputPath}'
|
||||
* SELECT (timestamp - ${startTime} AS relTime, ugi, cmd, src, dst, ip
|
||||
* FROM '${auditLogTableLocation}'
|
||||
* WHERE timestamp >= ${startTime} AND timestamp < ${endTime}
|
||||
* WHERE
|
||||
* timestamp {@literal >=} ${startTime}
|
||||
* AND timestamp {@literal <} ${endTime}
|
||||
* DISTRIBUTE BY src
|
||||
* SORT BY relTime ASC;
|
||||
* </pre>
|
||||
|
@ -57,16 +57,16 @@
|
||||
* format of these files is determined by the value of the
|
||||
* {@value COMMAND_PARSER_KEY} configuration, which defaults to
|
||||
* {@link AuditLogDirectParser}.
|
||||
*
|
||||
* <p/>This generates a number of {@link org.apache.hadoop.mapreduce.Counter}
|
||||
* <p>
|
||||
* This generates a number of {@link org.apache.hadoop.mapreduce.Counter}
|
||||
* values which can be used to get information into the replay, including the
|
||||
* number of commands replayed, how many of them were "invalid" (threw an
|
||||
* exception), how many were "late" (replayed later than they should have been),
|
||||
* and the latency (from client perspective) of each command. If there are a
|
||||
* large number of "late" commands, you likely need to increase the number of
|
||||
* threads used and/or the number of mappers.
|
||||
*
|
||||
* <p/>By default, commands will be replayed at the same rate as they were
|
||||
* <p>
|
||||
* By default, commands will be replayed at the same rate as they were
|
||||
* originally performed. However a rate factor can be specified via the
|
||||
* {@value RATE_FACTOR_KEY} configuration; all of the (relative) timestamps will
|
||||
* be divided by this rate factor, effectively changing the rate at which they
|
||||
|
Loading…
Reference in New Issue
Block a user