Apache
/
hadoop
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

YARN-4905. Improved "yarn logs" command-line to optionally show log metadata also. Contributed by Xuan Gong.

This commit is contained in:
Vinod Kumar Vavilapalli 2016-05-04 14:16:03 -07:00
parent 7bd418e48c
commit 9e37fe3b7a
4 changed files with 654 additions and 255 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

491
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-client/src/main/java/org/apache/hadoop/yarn/client/cli/LogsCLI.java
View File

@ -45,6 +45,7 @@
import org.apache.hadoop.util.Tool;
import org.apache.hadoop.yarn.api.records.ApplicationId;
import org.apache.hadoop.yarn.api.records.ApplicationReport;
import org.apache.hadoop.yarn.api.records.ContainerId;
import org.apache.hadoop.yarn.api.records.ContainerReport;
import org.apache.hadoop.yarn.api.records.YarnApplicationState;
import org.apache.hadoop.yarn.client.api.YarnClient;
@ -78,59 +79,16 @@ public class LogsCLI extends Configured implements Tool {
private static final String APP_OWNER_OPTION = "appOwner";
private static final String AM_CONTAINER_OPTION = "am";
private static final String CONTAINER_LOG_FILES = "logFiles";
private static final String SHOW_META_INFO = "show_meta_info";
private static final String LIST_NODES_OPTION = "list_nodes";
public static final String HELP_CMD = "help";
@Override
public int run(String[] args) throws Exception {
Options opts = new Options();
opts.addOption(HELP_CMD, false, "Displays help for all commands.");
Option appIdOpt =
new Option(APPLICATION_ID_OPTION, true, "ApplicationId (required)");
appIdOpt.setRequired(true);
opts.addOption(appIdOpt);
opts.addOption(CONTAINER_ID_OPTION, true, "ContainerId. "
+ "By default, it will only print syslog if the application is runing."
+ " Work with -logFiles to get other logs.");
opts.addOption(NODE_ADDRESS_OPTION, true, "NodeAddress in the format "
+ "nodename:port");
opts.addOption(APP_OWNER_OPTION, true,
"AppOwner (assumed to be current user if not specified)");
Option amOption = new Option(AM_CONTAINER_OPTION, true,
"Prints the AM Container logs for this application. "
+ "Specify comma-separated value to get logs for related AM Container. "
+ "For example, If we specify -am 1,2, we will get the logs for "
+ "the first AM Container as well as the second AM Container. "
+ "To get logs for all AM Containers, use -am ALL. "
+ "To get logs for the latest AM Container, use -am -1. "
+ "By default, it will only print out syslog. Work with -logFiles "
+ "to get other logs");
amOption.setValueSeparator(',');
amOption.setArgs(Option.UNLIMITED_VALUES);
amOption.setArgName("AM Containers");
opts.addOption(amOption);
Option logFileOpt = new Option(CONTAINER_LOG_FILES, true,
"Work with -am/-containerId and specify comma-separated value "
+ "to get specified container log files. Use \"ALL\" to fetch all the "
+ "log files for the container.");
logFileOpt.setValueSeparator(',');
logFileOpt.setArgs(Option.UNLIMITED_VALUES);
logFileOpt.setArgName("Log File Name");
opts.addOption(logFileOpt);
Options opts = createCommandOpts();
opts.getOption(APPLICATION_ID_OPTION).setArgName("Application ID");
opts.getOption(CONTAINER_ID_OPTION).setArgName("Container ID");
opts.getOption(NODE_ADDRESS_OPTION).setArgName("Node Address");
opts.getOption(APP_OWNER_OPTION).setArgName("Application Owner");
opts.getOption(AM_CONTAINER_OPTION).setArgName("AM Containers");
Options printOpts = new Options();
printOpts.addOption(opts.getOption(HELP_CMD));
printOpts.addOption(opts.getOption(CONTAINER_ID_OPTION));
printOpts.addOption(opts.getOption(NODE_ADDRESS_OPTION));
printOpts.addOption(opts.getOption(APP_OWNER_OPTION));
printOpts.addOption(opts.getOption(AM_CONTAINER_OPTION));
printOpts.addOption(opts.getOption(CONTAINER_LOG_FILES));
Options printOpts = createPrintOpts(opts);
if (args.length < 1) {
printHelpMessage(printOpts);
@ -146,6 +104,8 @@ public int run(String[] args) throws Exception {
String nodeAddress = null;
String appOwner = null;
boolean getAMContainerLogs = false;
boolean showMetaInfo = false;
boolean nodesList = false;
String[] logFiles = null;
List<String> amContainersList = new ArrayList<String>();
try {
@ -155,32 +115,15 @@ public int run(String[] args) throws Exception {
nodeAddress = commandLine.getOptionValue(NODE_ADDRESS_OPTION);
appOwner = commandLine.getOptionValue(APP_OWNER_OPTION);
getAMContainerLogs = commandLine.hasOption(AM_CONTAINER_OPTION);
showMetaInfo = commandLine.hasOption(SHOW_META_INFO);
nodesList = commandLine.hasOption(LIST_NODES_OPTION);
if (getAMContainerLogs) {
String[] amContainers = commandLine.getOptionValues(AM_CONTAINER_OPTION);
for (String am : amContainers) {
boolean errorInput = false;
if (!am.trim().equalsIgnoreCase("ALL")) {
try {
int id = Integer.parseInt(am.trim());
if (id != -1 && id <= 0) {
errorInput = true;
}
amContainersList = parseAMContainer(commandLine, printOpts);
} catch (NumberFormatException ex) {
errorInput = true;
}
if (errorInput) {
System.err.println(
"Invalid input for option -am. Valid inputs are 'ALL', -1 "
+ "and any other integer which is larger than 0.");
printHelpMessage(printOpts);
System.err.println(ex.getMessage());
return -1;
}
amContainersList.add(am.trim());
} else {
amContainersList.add("ALL");
break;
}
}
}
if (commandLine.hasOption(CONTAINER_LOG_FILES)) {
logFiles = commandLine.getOptionValues(CONTAINER_LOG_FILES);
@ -212,138 +155,54 @@ public int run(String[] args) throws Exception {
appOwner = UserGroupInformation.getCurrentUser().getShortUserName();
}
boolean appStateKnown = true;
boolean appStateObtainedSuccessfully = true;
YarnApplicationState appState = YarnApplicationState.NEW;
try {
appState = getApplicationState(appId);
if (appState == YarnApplicationState.NEW
|| appState == YarnApplicationState.NEW_SAVING
|| appState == YarnApplicationState.SUBMITTED) {
System.out.println("Logs are not avaiable right now.");
System.err.println("Logs are not avaiable right now.");
return -1;
}
} catch (IOException | YarnException e) {
appStateKnown = false;
appStateObtainedSuccessfully = false;
System.err.println("Unable to get ApplicationState."
+ " Attempting to fetch logs directly from the filesystem.");
}
if (showMetaInfo) {
return showMetaInfo(appState, appStateObtainedSuccessfully,
logCliHelper, appId, containerIdStr, nodeAddress, appOwner);
}
if (nodesList) {
return showNodeLists(appState, appStateObtainedSuccessfully,
logCliHelper, appId, appOwner);
}
// To get am logs
if (getAMContainerLogs) {
// if we do not specify the value for CONTAINER_LOG_FILES option,
// we will only output syslog
if (logFiles == null || logFiles.length == 0) {
logFiles = new String[] { "syslog" };
}
// If the application is running, we will call the RM WebService
// to get the AppAttempts which includes the nodeHttpAddress
// and containerId for all the AM Containers.
// After that, we will call NodeManager webService to get the
// related logs
if (appState == YarnApplicationState.ACCEPTED
|| appState == YarnApplicationState.RUNNING) {
return printAMContainerLogs(getConf(), appIdStr, amContainersList,
logFiles, logCliHelper, appOwner, false);
} else {
// If the application is in the final state, we will call RM webservice
// to get all AppAttempts information first. If we get nothing,
// we will try to call AHS webservice to get related AppAttempts
// which includes nodeAddress for the AM Containers.
// After that, we will use nodeAddress and containerId
// to get logs from HDFS directly.
if (getConf().getBoolean(YarnConfiguration.APPLICATION_HISTORY_ENABLED,
YarnConfiguration.DEFAULT_APPLICATION_HISTORY_ENABLED)) {
return printAMContainerLogs(getConf(), appIdStr, amContainersList,
logFiles, logCliHelper, appOwner, true);
} else {
System.out
.println(
"Can not get AMContainers logs for the application:" + appId);
System.out.println("This application:" + appId + " is finished."
+ " Please enable the application history service. Or Using "
+ "yarn logs -applicationId <appId> -containerId <containerId> "
+ "--nodeAddress <nodeHttpAddress> to get the container logs");
return -1;
}
}
return fetchAMContainerLogs(logFiles, appState, appId, appOwner,
amContainersList, logCliHelper);
}
int resultCode = 0;
if (containerIdStr != null) {
// if we provide the node address and the application is in the final
// state, we could directly get logs from HDFS.
if (nodeAddress != null && (!appStateKnown ||
isApplicationFinished(appState))) {
// if user specified "ALL" as the logFiles param, pass null
// to logCliHelper so that it fetches all the logs
List<String> logs;
if (logFiles == null) {
logs = null;
} else if (fetchAllLogFiles(logFiles)) {
logs = null;
} else {
logs = Arrays.asList(logFiles);
}
return logCliHelper.dumpAContainersLogsForALogType(appIdStr,
containerIdStr, nodeAddress, appOwner, logs);
}
String nodeHttpAddress = null;
String nodeId = null;
try {
// If the nodeAddress is not provided, we will try to get
// the ContainerReport. In the containerReport, we could get
// nodeAddress and nodeHttpAddress
ContainerReport report = getContainerReport(containerIdStr);
nodeHttpAddress =
report.getNodeHttpAddress().replaceFirst(
WebAppUtils.getHttpSchemePrefix(getConf()), "");
nodeId = report.getAssignedNode().toString();
} catch (IOException | YarnException ex) {
if (!appStateKnown || isApplicationFinished(appState)) {
String [] requestedLogFiles = logFiles;
if(fetchAllLogFiles(logFiles)) {
requestedLogFiles = null;
}
return printContainerLogsForFinishedApplicationWithoutNodeId(
appIdStr, containerIdStr, requestedLogFiles, logCliHelper,
appOwner);
} else if (!isApplicationFinished(appState)) {
System.err.println("Unable to get logs for this container:"
+ containerIdStr + "for the application:" + appId);
System.out.println("The application: " + appId + " is still running, "
+ "and we can not get Container report for the container: "
+ containerIdStr +". Please try later or after the application "
+ "finishes.");
ContainerId containerId = ContainerId.fromString(containerIdStr);
if (!containerId.getApplicationAttemptId().getApplicationId()
.equals(appId)) {
System.err.println("The Application:" + appId
+ " does not have the container:" + containerId);
return -1;
}
}
// If the application is not in the final state,
// we will provide the NodeHttpAddress and get the container logs
// by calling NodeManager webservice.
if (!isApplicationFinished(appState)) {
if (logFiles == null || logFiles.length == 0) {
logFiles = new String[] {"syslog"};
}
printContainerLogsFromRunningApplication(getConf(), appIdStr,
containerIdStr, nodeHttpAddress, nodeId, logFiles, logCliHelper,
appOwner);
} else {
String[] requestedLogFiles = logFiles;
if(fetchAllLogFiles(logFiles)) {
requestedLogFiles = null;
}
// If the application is in the final state, we will directly
// get the container logs from HDFS.
printContainerLogsForFinishedApplication(appIdStr, containerIdStr,
nodeId, requestedLogFiles, logCliHelper, appOwner);
}
return resultCode;
return fetchContainerLogs(appState, appStateObtainedSuccessfully,
logFiles, appOwner, nodeAddress, containerId, logCliHelper);
} else {
if (nodeAddress == null) {
resultCode =
logCliHelper.dumpAllContainersLogs(appId, appOwner, System.out);
} else {
System.out.println("Should at least provide ContainerId!");
System.err.println("Should at least provide ContainerId!");
printHelpMessage(printOpts);
resultCode = -1;
}
@ -382,7 +241,8 @@ public static void main(String[] args) throws Exception {
private void printHelpMessage(Options options) {
System.out.println("Retrieve logs for completed YARN applications.");
HelpFormatter formatter = new HelpFormatter();
formatter.printHelp("yarn logs -applicationId <application ID> [OPTIONS]", new Options());
formatter.printHelp("yarn logs -applicationId <application ID> [OPTIONS]",
new Options());
formatter.setSyntaxPrefix("");
formatter.printHelp("general options are:", options);
}
@ -410,9 +270,9 @@ private List<JSONObject> getAMContainerInfoForRMWebService(
return amContainersList;
}
private List<JSONObject> getAMContainerInfoForAHSWebService(Configuration conf,
String appId) throws ClientHandlerException, UniformInterfaceException,
JSONException {
private List<JSONObject> getAMContainerInfoForAHSWebService(
Configuration conf, String appId) throws ClientHandlerException,
UniformInterfaceException, JSONException {
Client webServiceClient = Client.create();
String webAppAddress =
WebAppUtils.getHttpSchemePrefix(conf)
@ -420,8 +280,9 @@ private List<JSONObject> getAMContainerInfoForAHSWebService(Configuration conf,
WebResource webResource = webServiceClient.resource(webAppAddress);
ClientResponse response =
webResource.path("ws").path("v1").path("applicationhistory").path("apps")
.path(appId).path("appattempts").accept(MediaType.APPLICATION_JSON)
webResource.path("ws").path("v1").path("applicationhistory")
.path("apps").path(appId).path("appattempts")
.accept(MediaType.APPLICATION_JSON)
.get(ClientResponse.class);
JSONObject json = response.getEntity(JSONObject.class);
JSONArray requests = json.getJSONArray("appAttempt");
@ -467,24 +328,27 @@ private String[] getContainerLogFiles(Configuration conf,
logFiles.add(elements.item(i).getTextContent());
}
} catch (Exception e) {
System.out.println("Unable to parse xml from webservice. Error:");
System.out.println(e.getMessage());
System.err.println("Unable to parse xml from webservice. Error:");
System.err.println(e.getMessage());
throw new IOException(e);
}
}
} catch (ClientHandlerException | UniformInterfaceException ex) {
System.out.println("Unable to fetch log files list");
System.err.println("Unable to fetch log files list");
throw new IOException(ex);
}
return logFiles.toArray(new String[0]);
}
private void printContainerLogsFromRunningApplication(Configuration conf,
String appId, String containerIdStr, String nodeHttpAddress,
ContainerId containerId, String nodeHttpAddress,
String nodeId, String[] logFiles, LogCLIHelpers logCliHelper,
String appOwner) throws IOException {
String [] requestedLogFiles = logFiles;
String appId = containerId.getApplicationAttemptId()
.getApplicationId().toString();
String containerIdStr = containerId.toString();
String[] requestedLogFiles = logFiles;
// fetch all the log files for the container
if (fetchAllLogFiles(logFiles)) {
requestedLogFiles =
@ -511,7 +375,7 @@ private void printContainerLogsFromRunningApplication(Configuration conf,
System.out.println(response.getEntity(String.class));
System.out.println("End of LogType:" + logFile);
} catch (ClientHandlerException | UniformInterfaceException ex) {
System.out.println("Can not find the log file:" + logFile
System.err.println("Can not find the log file:" + logFile
+ " for the container:" + containerIdStr + " in NodeManager:"
+ nodeId);
}
@ -528,7 +392,8 @@ private void printContainerLogsForFinishedApplication(String appId,
System.out.println(containerString);
System.out.println(StringUtils.repeat("=", containerString.length()));
logCliHelper.dumpAContainersLogsForALogType(appId, containerId,
nodeAddress, appOwner, logFiles != null ? Arrays.asList(logFiles) : null);
nodeAddress, appOwner, logFiles != null ? Arrays.asList(logFiles)
: null);
}
private int printContainerLogsForFinishedApplicationWithoutNodeId(
@ -648,7 +513,7 @@ private void outputAMContainerLogs(AMLogsRequest request, Configuration conf,
}
}
if (nodeId != null && !nodeId.isEmpty()) {
String [] requestedLogFilesList = null;
String[] requestedLogFilesList = null;
if(!fetchAllLogFiles(logFiles)) {
requestedLogFilesList = logFiles;
}
@ -659,18 +524,262 @@ private void outputAMContainerLogs(AMLogsRequest request, Configuration conf,
} else {
if (nodeHttpAddress != null && containerId != null
&& !nodeHttpAddress.isEmpty() && !containerId.isEmpty()) {
String [] requestedLogFiles = logFiles;
String[] requestedLogFiles = logFiles;
// fetch all the log files for the AM
if (fetchAllLogFiles(logFiles)) {
requestedLogFiles =
getContainerLogFiles(getConf(), containerId, nodeHttpAddress);
}
printContainerLogsFromRunningApplication(conf, appId, containerId,
nodeHttpAddress, nodeId, requestedLogFiles, logCliHelper, appOwner);
printContainerLogsFromRunningApplication(conf,
ContainerId.fromString(containerId), nodeHttpAddress, nodeId,
requestedLogFiles, logCliHelper, appOwner);
}
}
}
private int showMetaInfo(YarnApplicationState appState,
boolean appStateObtainedSuccessfully, LogCLIHelpers logCliHelper,
ApplicationId appId, String containerIdStr, String nodeAddress,
String appOwner) throws IOException {
if (!isApplicationFinished(appState) && appStateObtainedSuccessfully) {
System.err.println("The -show_meta_info command can be only used "
+ "with finished applications");
return -1;
} else {
logCliHelper.printLogMetadata(appId, containerIdStr, nodeAddress,
appOwner, System.out, System.err);
return 0;
}
}
private int showNodeLists(YarnApplicationState appState,
boolean appStateObtainedSuccessfully, LogCLIHelpers logCliHelper,
ApplicationId appId, String appOwner) throws IOException {
if (!isApplicationFinished(appState) && appStateObtainedSuccessfully) {
System.err.println("The -list_nodes command can be only used with "
+ "finished applications");
return -1;
} else {
logCliHelper.printNodesList(appId, appOwner, System.out, System.err);
return 0;
}
}
private Options createCommandOpts() {
Options opts = new Options();
opts.addOption(HELP_CMD, false, "Displays help for all commands.");
Option appIdOpt =
new Option(APPLICATION_ID_OPTION, true, "ApplicationId (required)");
appIdOpt.setRequired(true);
opts.addOption(appIdOpt);
opts.addOption(CONTAINER_ID_OPTION, true, "ContainerId. "
+ "By default, it will only print syslog if the application is runing."
+ " Work with -logFiles to get other logs.");
opts.addOption(NODE_ADDRESS_OPTION, true, "NodeAddress in the format "
+ "nodename:port");
opts.addOption(APP_OWNER_OPTION, true,
"AppOwner (assumed to be current user if not specified)");
Option amOption = new Option(AM_CONTAINER_OPTION, true,
"Prints the AM Container logs for this application. "
+ "Specify comma-separated value to get logs for related AM "
+ "Container. For example, If we specify -am 1,2, we will get "
+ "the logs for the first AM Container as well as the second "
+ "AM Container. To get logs for all AM Containers, use -am ALL. "
+ "To get logs for the latest AM Container, use -am -1. "
+ "By default, it will only print out syslog. Work with -logFiles "
+ "to get other logs");
amOption.setValueSeparator(',');
amOption.setArgs(Option.UNLIMITED_VALUES);
amOption.setArgName("AM Containers");
opts.addOption(amOption);
Option logFileOpt = new Option(CONTAINER_LOG_FILES, true,
"Work with -am/-containerId and specify comma-separated value "
+ "to get specified container log files. Use \"ALL\" to fetch all the "
+ "log files for the container.");
logFileOpt.setValueSeparator(',');
logFileOpt.setArgs(Option.UNLIMITED_VALUES);
logFileOpt.setArgName("Log File Name");
opts.addOption(logFileOpt);
opts.addOption(SHOW_META_INFO, false, "Show the log metadata, "
+ "including log-file names, the size of the log files. "
+ "You can combine this with --containerId to get log metadata for "
+ "the specific container, or with --nodeAddress to get log metadata "
+ "for all the containers on the specific NodeManager. "
+ "Currently, this option can only be used for finished "
+ "applications.");
opts.addOption(LIST_NODES_OPTION, false,
"Show the list of nodes that successfully aggregated logs. "
+ "This option can only be used with finished applications.");
opts.getOption(APPLICATION_ID_OPTION).setArgName("Application ID");
opts.getOption(CONTAINER_ID_OPTION).setArgName("Container ID");
opts.getOption(NODE_ADDRESS_OPTION).setArgName("Node Address");
opts.getOption(APP_OWNER_OPTION).setArgName("Application Owner");
opts.getOption(AM_CONTAINER_OPTION).setArgName("AM Containers");
return opts;
}
private Options createPrintOpts(Options commandOpts) {
Options printOpts = new Options();
printOpts.addOption(commandOpts.getOption(HELP_CMD));
printOpts.addOption(commandOpts.getOption(CONTAINER_ID_OPTION));
printOpts.addOption(commandOpts.getOption(NODE_ADDRESS_OPTION));
printOpts.addOption(commandOpts.getOption(APP_OWNER_OPTION));
printOpts.addOption(commandOpts.getOption(AM_CONTAINER_OPTION));
printOpts.addOption(commandOpts.getOption(CONTAINER_LOG_FILES));
printOpts.addOption(commandOpts.getOption(SHOW_META_INFO));
printOpts.addOption(commandOpts.getOption(LIST_NODES_OPTION));
return printOpts;
}
private List<String> parseAMContainer(CommandLine commandLine,
Options printOpts) throws NumberFormatException {
List<String> amContainersList = new ArrayList<String>();
String[] amContainers = commandLine.getOptionValues(AM_CONTAINER_OPTION);
for (String am : amContainers) {
boolean errorInput = false;
if (!am.trim().equalsIgnoreCase("ALL")) {
try {
int id = Integer.parseInt(am.trim());
if (id != -1 && id <= 0) {
errorInput = true;
}
} catch (NumberFormatException ex) {
errorInput = true;
}
if (errorInput) {
String errMessage =
"Invalid input for option -am. Valid inputs are 'ALL', -1 "
+ "and any other integer which is larger than 0.";
printHelpMessage(printOpts);
throw new NumberFormatException(errMessage);
}
amContainersList.add(am.trim());
} else {
amContainersList.add("ALL");
break;
}
}
return amContainersList;
}
private int fetchAMContainerLogs(String[] logFiles,
YarnApplicationState appState, ApplicationId appId,
String appOwner, List<String> amContainersList,
LogCLIHelpers logCliHelper) throws Exception {
// if we do not specify the value for CONTAINER_LOG_FILES option,
// we will only output syslog
if (logFiles == null || logFiles.length == 0) {
logFiles = new String[] {"syslog"};
}
// If the application is running, we will call the RM WebService
// to get the AppAttempts which includes the nodeHttpAddress
// and containerId for all the AM Containers.
// After that, we will call NodeManager webService to get the
// related logs
if (appState == YarnApplicationState.ACCEPTED
|| appState == YarnApplicationState.RUNNING) {
return printAMContainerLogs(getConf(), appId.toString(), amContainersList,
logFiles, logCliHelper, appOwner, false);
} else {
// If the application is in the final state, we will call RM webservice
// to get all AppAttempts information first. If we get nothing,
// we will try to call AHS webservice to get related AppAttempts
// which includes nodeAddress for the AM Containers.
// After that, we will use nodeAddress and containerId
// to get logs from HDFS directly.
if (getConf().getBoolean(YarnConfiguration.APPLICATION_HISTORY_ENABLED,
YarnConfiguration.DEFAULT_APPLICATION_HISTORY_ENABLED)) {
return printAMContainerLogs(getConf(), appId.toString(),
amContainersList, logFiles, logCliHelper, appOwner, true);
} else {
System.err.println("Can not get AMContainers logs for "
+ "the application:" + appId);
System.err.println("This application:" + appId + " is finished."
+ " Please enable the application history service. Or Using "
+ "yarn logs -applicationId <appId> -containerId <containerId> "
+ "--nodeAddress <nodeHttpAddress> to get the container logs");
return -1;
}
}
}
private int fetchContainerLogs(YarnApplicationState appState,
boolean appStateObtainedSuccessfully, String[] logFiles,
String appOwner, String nodeAddress,
ContainerId containerId, LogCLIHelpers logCliHelper) throws IOException {
int resultCode = 0;
String appIdStr = containerId.getApplicationAttemptId()
.getApplicationId().toString();
String containerIdStr = containerId.toString();
// if we provide the node address and the application is in the final
// state, we could directly get logs from HDFS.
if (nodeAddress != null && (!appStateObtainedSuccessfully ||
isApplicationFinished(appState))) {
// if user specified "ALL" as the logFiles param, pass null
// to logCliHelper so that it fetches all the logs
List<String> logs;
if (logFiles == null) {
logs = null;
} else if (fetchAllLogFiles(logFiles)) {
logs = null;
} else {
logs = Arrays.asList(logFiles);
}
return logCliHelper.dumpAContainersLogsForALogType(appIdStr,
containerIdStr, nodeAddress, appOwner, logs);
}
String nodeHttpAddress = null;
String nodeId = null;
try {
// If the nodeAddress is not provided, we will try to get
// the ContainerReport. In the containerReport, we could get
// nodeAddress and nodeHttpAddress
ContainerReport report = getContainerReport(containerIdStr);
nodeHttpAddress =
report.getNodeHttpAddress().replaceFirst(
WebAppUtils.getHttpSchemePrefix(getConf()), "");
nodeId = report.getAssignedNode().toString();
} catch (IOException | YarnException ex) {
if (!appStateObtainedSuccessfully || isApplicationFinished(appState)) {
String[] requestedLogFiles = logFiles;
if(fetchAllLogFiles(logFiles)) {
requestedLogFiles = null;
}
return printContainerLogsForFinishedApplicationWithoutNodeId(
appIdStr, containerIdStr, requestedLogFiles, logCliHelper,
appOwner);
} else if (!isApplicationFinished(appState)) {
System.err.println("Unable to get logs for this container:"
+ containerIdStr + "for the application:" + appIdStr);
System.err.println("The application: " + appIdStr
+ " is still running, and we can not get Container report "
+ "for the container: " + containerIdStr +". Please try later "
+ "or after the application finishes.");
return -1;
}
}
// If the application is not in the final state,
// we will provide the NodeHttpAddress and get the container logs
// by calling NodeManager webservice.
if (!isApplicationFinished(appState)) {
if (logFiles == null || logFiles.length == 0) {
logFiles = new String[] {"syslog"};
}
printContainerLogsFromRunningApplication(getConf(), containerId,
nodeHttpAddress, nodeId, logFiles, logCliHelper, appOwner);
} else {
String[] requestedLogFiles = logFiles;
if(fetchAllLogFiles(logFiles)) {
requestedLogFiles = null;
}
// If the application is in the final state, we will directly
// get the container logs from HDFS.
printContainerLogsForFinishedApplication(appIdStr, containerIdStr,
nodeId, requestedLogFiles, logCliHelper, appOwner);
}
return resultCode;
}
private static class AMLogsRequest {
private String amContainerId;
private String nodeId;

190
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-client/src/test/java/org/apache/hadoop/yarn/client/cli/TestLogsCLI.java
View File

@ -18,6 +18,7 @@
package org.apache.hadoop.yarn.client.cli;
import static org.junit.Assert.assertFalse;
import static org.junit.Assert.assertNotNull;
import static org.junit.Assert.assertTrue;
import static org.mockito.Matchers.any;
@ -166,11 +167,22 @@ public void testHelpMessage() throws Exception {
pw.println(" runing. Work with -logFiles to get other");
pw.println(" logs.");
pw.println(" -help Displays help for all commands.");
pw.println(" -list_nodes Show the list of nodes that successfully");
pw.println(" aggregated logs. This option can only be");
pw.println(" used with finished applications.");
pw.println(" -logFiles <Log File Name> Work with -am/-containerId and specify");
pw.println(" comma-separated value to get specified");
pw.println(" container log files. Use \"ALL\" to fetch");
pw.println(" all the log files for the container.");
pw.println(" -nodeAddress <Node Address> NodeAddress in the format nodename:port");
pw.println(" -show_meta_info Show the log metadata, including log-file");
pw.println(" names, the size of the log files. You can");
pw.println(" combine this with --containerId to get");
pw.println(" log metadata for the specific container,");
pw.println(" or with --nodeAddress to get log metadata");
pw.println(" for all the containers on the specific");
pw.println(" NodeManager. Currently, this option can");
pw.println(" only be used for finished applications.");
pw.close();
String appReportStr = baos.toString("UTF-8");
Assert.assertEquals(appReportStr, sysOutStream.toString());
@ -286,10 +298,10 @@ public void testFetchApplictionLogs() throws Exception {
"-nodeAddress", nodeId.toString(), "-containerId",
containerId0.toString() });
assertTrue(exitCode == -1);
assertTrue(sysOutStream.toString().contains(
assertTrue(sysErrStream.toString().contains(
"Logs for container " + containerId0.toString()
+ " are not present in this log-file."));
sysOutStream.reset();
sysErrStream.reset();
// uploaded two logs for container3. The first log is named as syslog.
// The second one is named as stdout.
@ -336,6 +348,140 @@ public void testFetchApplictionLogs() throws Exception {
fs.delete(new Path(rootLogDir), true);
}
@Test (timeout = 15000)
public void testPrintContainerLogMetadata() throws Exception {
String remoteLogRootDir = "target/logs/";
Configuration configuration = new Configuration();
configuration.setBoolean(YarnConfiguration.LOG_AGGREGATION_ENABLED, true);
configuration
.set(YarnConfiguration.NM_REMOTE_APP_LOG_DIR, remoteLogRootDir);
configuration.setBoolean(YarnConfiguration.YARN_ACL_ENABLE, true);
configuration.set(YarnConfiguration.YARN_ADMIN_ACL, "admin");
FileSystem fs = FileSystem.get(configuration);
String rootLogDir = "target/LocalLogs";
ApplicationId appId = ApplicationId.newInstance(0, 1);
ApplicationAttemptId appAttemptId =
ApplicationAttemptId.newInstance(appId, 1);
List<ContainerId> containerIds = new ArrayList<ContainerId>();
ContainerId containerId1 = ContainerId.newContainerId(
appAttemptId, 1);
ContainerId containerId2 = ContainerId.newContainerId(
appAttemptId, 2);
containerIds.add(containerId1);
containerIds.add(containerId2);
List<NodeId> nodeIds = new ArrayList<NodeId>();
NodeId nodeId = NodeId.newInstance("localhost", 1234);
nodeIds.add(nodeId);
nodeIds.add(nodeId);
createContainerLogs(configuration, remoteLogRootDir, rootLogDir, fs,
appId, containerIds, nodeIds);
YarnClient mockYarnClient =
createMockYarnClient(YarnApplicationState.FINISHED);
LogsCLI cli = new LogsCLIForTest(mockYarnClient);
cli.setConf(configuration);
cli.run(new String[] { "-applicationId", appId.toString(),
"-show_meta_info" });
assertTrue(sysOutStream.toString().contains(
"Container: container_0_0001_01_000001 on localhost_"));
assertTrue(sysOutStream.toString().contains(
"Container: container_0_0001_01_000002 on localhost_"));
assertTrue(sysOutStream.toString().contains(
"LogType:syslog"));
assertTrue(sysOutStream.toString().contains(
"LogLength:43"));
sysOutStream.reset();
cli.run(new String[] { "-applicationId", appId.toString(),
"-show_meta_info", "-containerId", "container_0_0001_01_000001" });
assertTrue(sysOutStream.toString().contains(
"Container: container_0_0001_01_000001 on localhost_"));
assertFalse(sysOutStream.toString().contains(
"Container: container_0_0001_01_000002 on localhost_"));
assertTrue(sysOutStream.toString().contains(
"LogType:syslog"));
assertTrue(sysOutStream.toString().contains(
"LogLength:43"));
sysOutStream.reset();
cli.run(new String[] { "-applicationId", appId.toString(),
"-show_meta_info", "-nodeAddress", "localhost" });
assertTrue(sysOutStream.toString().contains(
"Container: container_0_0001_01_000001 on localhost_"));
assertTrue(sysOutStream.toString().contains(
"Container: container_0_0001_01_000002 on localhost_"));
assertTrue(sysOutStream.toString().contains(
"LogType:syslog"));
assertTrue(sysOutStream.toString().contains(
"LogLength:43"));
sysOutStream.reset();
cli.run(new String[] { "-applicationId", appId.toString(),
"-show_meta_info", "-nodeAddress", "localhost", "-containerId",
"container_1234" });
assertTrue(sysErrStream.toString().contains(
"The container container_1234 couldn't be found on the node "
+ "specified: localhost"));
sysErrStream.reset();
fs.delete(new Path(remoteLogRootDir), true);
fs.delete(new Path(rootLogDir), true);
}
@Test (timeout = 15000)
public void testListNodeInfo() throws Exception {
String remoteLogRootDir = "target/logs/";
Configuration configuration = new Configuration();
configuration.setBoolean(YarnConfiguration.LOG_AGGREGATION_ENABLED, true);
configuration
.set(YarnConfiguration.NM_REMOTE_APP_LOG_DIR, remoteLogRootDir);
configuration.setBoolean(YarnConfiguration.YARN_ACL_ENABLE, true);
configuration.set(YarnConfiguration.YARN_ADMIN_ACL, "admin");
ApplicationId appId = ApplicationId.newInstance(0, 1);
ApplicationAttemptId appAttemptId =
ApplicationAttemptId.newInstance(appId, 1);
List<ContainerId> containerIds = new ArrayList<ContainerId>();
ContainerId containerId1 = ContainerId.newContainerId(
appAttemptId, 1);
ContainerId containerId2 = ContainerId.newContainerId(
appAttemptId, 2);
containerIds.add(containerId1);
containerIds.add(containerId2);
List<NodeId> nodeIds = new ArrayList<NodeId>();
NodeId nodeId1 = NodeId.newInstance("localhost1", 1234);
NodeId nodeId2 = NodeId.newInstance("localhost2", 2345);
nodeIds.add(nodeId1);
nodeIds.add(nodeId2);
String rootLogDir = "target/LocalLogs";
FileSystem fs = FileSystem.get(configuration);
createContainerLogs(configuration, remoteLogRootDir, rootLogDir, fs,
appId, containerIds, nodeIds);
YarnClient mockYarnClient =
createMockYarnClient(YarnApplicationState.FINISHED);
LogsCLI cli = new LogsCLIForTest(mockYarnClient);
cli.setConf(configuration);
cli.run(new String[] { "-applicationId", appId.toString(),
"-list_nodes" });
assertTrue(sysOutStream.toString().contains(
LogAggregationUtils.getNodeString(nodeId1)));
assertTrue(sysOutStream.toString().contains(
LogAggregationUtils.getNodeString(nodeId2)));
sysOutStream.reset();
fs.delete(new Path(remoteLogRootDir), true);
fs.delete(new Path(rootLogDir), true);
}
@Test (timeout = 15000)
public void testFetchApplictionLogsHar() throws Exception {
String remoteLogRootDir = "target/logs/";
@ -384,6 +530,46 @@ public void testFetchApplictionLogsHar() throws Exception {
fs.delete(new Path(remoteLogRootDir), true);
}
private void createContainerLogs(Configuration configuration,
String remoteLogRootDir, String rootLogDir, FileSystem fs,
ApplicationId appId, List<ContainerId> containerIds,
List<NodeId> nodeIds) throws Exception {
UserGroupInformation ugi = UserGroupInformation.getCurrentUser();
// create local logs
Path rootLogDirPath = new Path(rootLogDir);
if (fs.exists(rootLogDirPath)) {
fs.delete(rootLogDirPath, true);
}
assertTrue(fs.mkdirs(rootLogDirPath));
Path appLogsDir = new Path(rootLogDirPath, appId.toString());
if (fs.exists(appLogsDir)) {
fs.delete(appLogsDir, true);
}
assertTrue(fs.mkdirs(appLogsDir));
List<String> rootLogDirs = Arrays.asList(rootLogDir);
List<String> logTypes = new ArrayList<String>();
logTypes.add("syslog");
// create container logs in localLogDir
for (ContainerId containerId : containerIds) {
createContainerLogInLocalDir(appLogsDir, containerId, fs, logTypes);
}
Path path =
new Path(remoteLogRootDir + ugi.getShortUserName()
+ "/logs/application_0_0001");
if (fs.exists(path)) {
fs.delete(path, true);
}
assertTrue(fs.mkdirs(path));
for (int i=0; i<containerIds.size(); i++) {
uploadContainerLogIntoRemoteDir(ugi, configuration, rootLogDirs, nodeIds.get(i),
containerIds.get(i), path, fs);
}
}
private static void createContainerLogInLocalDir(Path appLogsDir,
ContainerId containerId, FileSystem fs, List<String> logTypes) throws Exception {
Path containerLogsDir = new Path(appLogsDir, containerId.toString());

20
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common/src/main/java/org/apache/hadoop/yarn/logaggregation/AggregatedLogFormat.java
View File

@ -808,6 +808,26 @@ public static int readContainerLogsForALogType(
}
}
@Private
public static void readContainerMetaDataAndSkipData(
DataInputStream valueStream, PrintStream out) throws IOException {
String fileType = valueStream.readUTF();
String fileLengthStr = valueStream.readUTF();
long fileLength = Long.parseLong(fileLengthStr);
out.print("LogType:");
out.println(fileType);
out.print("LogLength:");
out.println(fileLengthStr);
long totalSkipped = 0;
long currSkipped = 0;
while (currSkipped != -1 && totalSkipped < fileLength) {
currSkipped = valueStream.skip(fileLength - totalSkipped);
totalSkipped += currSkipped;
}
}
public void close() {
IOUtils.cleanup(LOG, scanner, reader, fsDataIStream);
}

192
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common/src/main/java/org/apache/hadoop/yarn/logaggregation/LogCLIHelpers.java
View File

@ -68,24 +68,10 @@ public int dumpAContainersLogsForALogType(String appId, String containerId,
public int dumpAContainersLogsForALogType(String appId, String containerId,
String nodeId, String jobOwner, List<String> logType,
boolean outputFailure) throws IOException {
Path remoteRootLogDir = new Path(getConf().get(
YarnConfiguration.NM_REMOTE_APP_LOG_DIR,
YarnConfiguration.DEFAULT_NM_REMOTE_APP_LOG_DIR));
String suffix = LogAggregationUtils.getRemoteNodeLogDirSuffix(getConf());
ApplicationId applicationId = ConverterUtils.toApplicationId(appId);
Path remoteAppLogDir = LogAggregationUtils.getRemoteAppLogDir(
remoteRootLogDir, applicationId, jobOwner,
suffix);
RemoteIterator<FileStatus> nodeFiles;
try {
Path qualifiedLogDir =
FileContext.getFileContext(getConf()).makeQualified(
remoteAppLogDir);
nodeFiles =
FileContext.getFileContext(qualifiedLogDir.toUri(), getConf())
.listStatus(remoteAppLogDir);
} catch (FileNotFoundException fnf) {
logDirNotExist(remoteAppLogDir.toString());
RemoteIterator<FileStatus> nodeFiles = getRemoteNodeFileDir(
applicationId, jobOwner);
if (nodeFiles == null) {
return -1;
}
boolean foundContainerLogs = false;
@ -134,23 +120,10 @@ public int dumpAContainersLogsForALogType(String appId, String containerId,
public int dumpAContainersLogsForALogTypeWithoutNodeId(String appId,
String containerId, String jobOwner, List<String> logType)
throws IOException {
Path remoteRootLogDir = new Path(getConf().get(
YarnConfiguration.NM_REMOTE_APP_LOG_DIR,
YarnConfiguration.DEFAULT_NM_REMOTE_APP_LOG_DIR));
ApplicationId applicationId = ConverterUtils.toApplicationId(appId);
String user = jobOwner;
String logDirSuffix = LogAggregationUtils.getRemoteNodeLogDirSuffix(
getConf());
Path remoteAppLogDir = LogAggregationUtils.getRemoteAppLogDir(
remoteRootLogDir, applicationId, user, logDirSuffix);
RemoteIterator<FileStatus> nodeFiles;
try {
Path qualifiedLogDir =
FileContext.getFileContext(getConf()).makeQualified(remoteAppLogDir);
nodeFiles = FileContext.getFileContext(qualifiedLogDir.toUri(),
getConf()).listStatus(remoteAppLogDir);
} catch (FileNotFoundException fnf) {
logDirNotExist(remoteAppLogDir.toString());
RemoteIterator<FileStatus> nodeFiles = getRemoteNodeFileDir(
applicationId, jobOwner);
if (nodeFiles == null) {
return -1;
}
boolean foundContainerLogs = false;
@ -261,22 +234,9 @@ public int dumpAContainerLogsForALogType(String containerIdStr,
@Private
public int dumpAllContainersLogs(ApplicationId appId, String appOwner,
PrintStream out) throws IOException {
Path remoteRootLogDir = new Path(getConf().get(
YarnConfiguration.NM_REMOTE_APP_LOG_DIR,
YarnConfiguration.DEFAULT_NM_REMOTE_APP_LOG_DIR));
String user = appOwner;
String logDirSuffix = LogAggregationUtils.getRemoteNodeLogDirSuffix(getConf());
// TODO Change this to get a list of files from the LAS.
Path remoteAppLogDir = LogAggregationUtils.getRemoteAppLogDir(
remoteRootLogDir, appId, user, logDirSuffix);
RemoteIterator<FileStatus> nodeFiles;
try {
Path qualifiedLogDir =
FileContext.getFileContext(getConf()).makeQualified(remoteAppLogDir);
nodeFiles = FileContext.getFileContext(qualifiedLogDir.toUri(),
getConf()).listStatus(remoteAppLogDir);
} catch (FileNotFoundException fnf) {
logDirNotExist(remoteAppLogDir.toString());
RemoteIterator<FileStatus> nodeFiles = getRemoteNodeFileDir(
appId, appOwner);
if (nodeFiles == null) {
return -1;
}
boolean foundAnyLogs = false;
@ -324,12 +284,136 @@ public int dumpAllContainersLogs(ApplicationId appId, String appOwner,
}
}
if (! foundAnyLogs) {
emptyLogDir(remoteAppLogDir.toString());
emptyLogDir(getRemoteAppLogDir(appId, appOwner).toString());
return -1;
}
return 0;
}
@Private
public void printLogMetadata(ApplicationId appId,
String containerIdStr, String nodeId, String appOwner,
PrintStream out, PrintStream err)
throws IOException {
boolean getAllContainers = (containerIdStr == null);
String nodeIdStr = (nodeId == null) ? null
: LogAggregationUtils.getNodeString(nodeId);
RemoteIterator<FileStatus> nodeFiles = getRemoteNodeFileDir(
appId, appOwner);
if (nodeFiles == null) {
return;
}
boolean foundAnyLogs = false;
while (nodeFiles.hasNext()) {
FileStatus thisNodeFile = nodeFiles.next();
if (nodeIdStr != null) {
if (!thisNodeFile.getPath().getName().contains(nodeIdStr)) {
continue;
}
}
if (!thisNodeFile.getPath().getName()
.endsWith(LogAggregationUtils.TMP_FILE_SUFFIX)) {
AggregatedLogFormat.LogReader reader =
new AggregatedLogFormat.LogReader(getConf(),
thisNodeFile.getPath());
try {
DataInputStream valueStream;
LogKey key = new LogKey();
valueStream = reader.next(key);
while (valueStream != null) {
if (getAllContainers || (key.toString().equals(containerIdStr))) {
String containerString =
"\n\nContainer: " + key + " on "
+ thisNodeFile.getPath().getName();
out.println(containerString);
out.println("Log Upload Time:"
+ thisNodeFile.getModificationTime());
out.println(StringUtils.repeat("=", containerString.length()));
while (true) {
try {
LogReader.readContainerMetaDataAndSkipData(valueStream, out);
} catch (EOFException eof) {
break;
}
}
foundAnyLogs = true;
if (!getAllContainers) {
break;
}
}
// Next container
key = new LogKey();
valueStream = reader.next(key);
}
} finally {
reader.close();
}
}
}
if (!foundAnyLogs) {
if (containerIdStr != null && nodeId != null) {
err.println("The container " + containerIdStr + " couldn't be found "
+ "on the node specified: " + nodeId);
} else if (nodeId != null) {
err.println("Can not find log metadata for any containers on "
+ nodeId);
} else if (containerIdStr != null) {
err.println("Can not find log metadata for container: "
+ containerIdStr);
}
}
}
@Private
public void printNodesList(ApplicationId appId, String appOwner,
PrintStream out, PrintStream err) throws IOException {
RemoteIterator<FileStatus> nodeFiles = getRemoteNodeFileDir(
appId, appOwner);
if (nodeFiles == null) {
return;
}
boolean foundNode = false;
StringBuilder sb = new StringBuilder();
while (nodeFiles.hasNext()) {
FileStatus thisNodeFile = nodeFiles.next();
sb.append(thisNodeFile.getPath().getName() + "\n");
foundNode = true;
}
if (!foundNode) {
err.println("No nodes found that aggregated logs for "
+ "the application: " + appId);
} else {
out.println(sb.toString());
}
}
private RemoteIterator<FileStatus> getRemoteNodeFileDir(ApplicationId appId,
String appOwner) throws IOException {
Path remoteAppLogDir = getRemoteAppLogDir(appId, appOwner);
RemoteIterator<FileStatus> nodeFiles = null;
try {
Path qualifiedLogDir =
FileContext.getFileContext(getConf()).makeQualified(remoteAppLogDir);
nodeFiles = FileContext.getFileContext(qualifiedLogDir.toUri(),
getConf()).listStatus(remoteAppLogDir);
} catch (FileNotFoundException fnf) {
logDirNotExist(remoteAppLogDir.toString());
}
return nodeFiles;
}
private Path getRemoteAppLogDir(ApplicationId appId, String appOwner) {
Path remoteRootLogDir = new Path(getConf().get(
YarnConfiguration.NM_REMOTE_APP_LOG_DIR,
YarnConfiguration.DEFAULT_NM_REMOTE_APP_LOG_DIR));
String user = appOwner;
String logDirSuffix = LogAggregationUtils
.getRemoteNodeLogDirSuffix(getConf());
// TODO Change this to get a list of files from the LAS.
return LogAggregationUtils.getRemoteAppLogDir(
remoteRootLogDir, appId, user, logDirSuffix);
}
@Override
public void setConf(Configuration conf) {
this.conf = conf;
@ -341,16 +425,16 @@ public Configuration getConf() {
}
private static void containerLogNotFound(String containerId) {
System.out.println("Logs for container " + containerId
System.err.println("Logs for container " + containerId
+ " are not present in this log-file.");
}
private static void logDirNotExist(String remoteAppLogDir) {
System.out.println(remoteAppLogDir + " does not exist.");
System.out.println("Log aggregation has not completed or is not enabled.");
System.err.println(remoteAppLogDir + " does not exist.");
System.err.println("Log aggregation has not completed or is not enabled.");
}
private static void emptyLogDir(String remoteAppLogDir) {
System.out.println(remoteAppLogDir + " does not have any log files.");
System.err.println(remoteAppLogDir + " does not have any log files.");
}
}