/**
* 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.hadoop.mapred;
import java.io.File;
import java.io.IOException;
import java.io.OutputStream;
import java.net.URL;
import java.net.HttpURLConnection;
import java.net.URLEncoder;
import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import org.apache.hadoop.fs.CommonConfigurationKeys;
import org.apache.hadoop.fs.FileUtil;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.http.TestHttpServer.DummyFilterInitializer;
import org.apache.hadoop.mapred.JobHistory.Keys;
import org.apache.hadoop.mapred.JobHistory.TaskAttempt;
import org.apache.hadoop.mapred.QueueManager.QueueACL;
import org.apache.hadoop.mapreduce.JobContext;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.examples.SleepJob;
import org.apache.hadoop.security.Groups;
import org.apache.hadoop.security.ShellBasedUnixGroupsMapping;
import org.apache.hadoop.security.UserGroupInformation;
import org.junit.Test;
import java.security.PrivilegedExceptionAction;
import java.util.Arrays;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Properties;
public class TestWebUIAuthorization extends ClusterMapReduceTestCase {
private static final Log LOG = LogFactory.getLog(
TestWebUIAuthorization.class);
// users who submit the jobs
private static final String jobSubmitter = "user1";
private static final String jobSubmitter1 = "user11";
private static final String jobSubmitter2 = "user12";
private static final String jobSubmitter3 = "user13";
// mrOwner starts the cluster
private static String mrOwner = null;
// admin of "default" queue
private static final String qAdmin = "user3";
// mrAdmin
private static final String mrAdminUser = "user4";
// Group for mrAdmins
private static final String mrAdminGroup = "admingroup";
// member of mrAdminGroup
private static final String mrAdminGroupMember = "user2";
// "colleague1" is there in job-view-acls config
private static final String viewColleague = "colleague1";
// "colleague2" is there in job-modify-acls config
private static final String modifyColleague = "colleague2";
// "colleague3" is there in both job-view-acls and job-modify-acls
private static final String viewAndModifyColleague = "colleague3";
// "evilJohn" is not having view/modify access on the jobs
private static final String unauthorizedUser = "evilJohn";
protected void setUp() throws Exception {
// do not do anything
};
/** access a url, ignoring some IOException such as the page does not exist */
static int getHttpStatusCode(String urlstring, String userName,
String method) throws IOException {
LOG.info("Accessing " + urlstring + " as user " + userName);
URL url = new URL(urlstring + "&user.name=" + userName);
HttpURLConnection connection = (HttpURLConnection)url.openConnection();
connection.setRequestMethod(method);
if (method.equals("POST")) {
String encodedData = "action=kill&user.name=" + userName;
connection.setRequestProperty("Content-Type",
"application/x-www-form-urlencoded");
connection.setRequestProperty("Content-Length",
Integer.toString(encodedData.length()));
connection.setDoOutput(true);
OutputStream os = connection.getOutputStream();
os.write(encodedData.getBytes());
}
connection.connect();
return connection.getResponseCode();
}
public static class MyGroupsProvider extends ShellBasedUnixGroupsMapping {
static Map<String, List<String>> mapping = new HashMap<String, List<String>>();
@Override
public List<String> getGroups(String user) throws IOException {
return mapping.get(user);
}
}
/**
* Validates the given jsp/servlet against different user names who
* can(or cannot) view the job.
* (1) jobSubmitter can view the job
* (2) mrAdmin can view any job
* (3) mrOwner can view any job
* (4) qAdmins of the queue to which job is submitted to can view any job in
* that queue.
* (5) user mentioned in job-view-acl should be able to view the
* job irrespective of job-modify-acl.
* (6) user mentioned in job-modify-acl but not in job-view-acl
* cannot view the job
* (7) other unauthorized users cannot view the job
*/
private void validateViewJob(String url, String method)
throws IOException {
assertEquals("Incorrect return code for job submitter " + jobSubmitter,
HttpURLConnection.HTTP_OK, getHttpStatusCode(url, jobSubmitter,
method));
assertEquals("Incorrect return code for admin user " +
mrAdminUser, HttpURLConnection.HTTP_OK,
getHttpStatusCode(url, mrAdminUser, method));
assertEquals("Incorrect return code for admingroup-member " +
mrAdminGroupMember, HttpURLConnection.HTTP_OK,
getHttpStatusCode(url, mrAdminGroupMember, method));
assertEquals("Incorrect return code for MR-owner " + mrOwner,
HttpURLConnection.HTTP_OK, getHttpStatusCode(url, mrOwner, method));
assertEquals("Incorrect return code for queue admin " + qAdmin,
HttpURLConnection.HTTP_OK, getHttpStatusCode(url, qAdmin, method));
assertEquals("Incorrect return code for user in job-view-acl " +
viewColleague, HttpURLConnection.HTTP_OK,
getHttpStatusCode(url, viewColleague, method));
assertEquals("Incorrect return code for user in job-view-acl and " +
"job-modify-acl " + viewAndModifyColleague, HttpURLConnection.HTTP_OK,
getHttpStatusCode(url, viewAndModifyColleague, method));
assertEquals("Incorrect return code for user in job-modify-acl " +
modifyColleague, HttpURLConnection.HTTP_UNAUTHORIZED,
getHttpStatusCode(url, modifyColleague, method));
assertEquals("Incorrect return code for unauthorizedUser " +
unauthorizedUser, HttpURLConnection.HTTP_UNAUTHORIZED,
getHttpStatusCode(url, unauthorizedUser, method));
}
/**
* Validates the given jsp/servlet against different user names who
* can(or cannot) modify the job.
* (1) jobSubmitter, mrOwner, qAdmin and mrAdmin can modify the job.
* But we are not validating this in this method. Let the caller
* explicitly validate this, if needed.
* (2) user mentioned in job-view-acl but not in job-modify-acl cannot
* modify the job
* (3) user mentioned in job-modify-acl (irrespective of job-view-acl)
* can modify the job
* (4) other unauthorized users cannot modify the job
*/
private void validateModifyJob(String url, String method) throws IOException {
assertEquals(HttpURLConnection.HTTP_UNAUTHORIZED,
getHttpStatusCode(url, viewColleague, method));
assertEquals(HttpURLConnection.HTTP_UNAUTHORIZED,
getHttpStatusCode(url, unauthorizedUser, method));
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(url, modifyColleague, method));
}
// starts a sleep job with 1 map task that runs for a long time
private RunningJob startSleepJobAsUser(String user, final JobConf conf,
MiniMRCluster cluster) throws Exception {
final SleepJob sleepJob = new SleepJob();
sleepJob.setConf(conf);
UserGroupInformation jobSubmitterUGI =
UserGroupInformation.createRemoteUser(user);
RunningJob job =
jobSubmitterUGI.doAs(new PrivilegedExceptionAction<RunningJob>() {
public RunningJob run() throws Exception {
JobClient jobClient = new JobClient(conf);
SleepJob sleepJob = new SleepJob();
sleepJob.setConf(conf);
JobConf jobConf =
sleepJob.setupJobConf(1, 0, 900000, 1, 1000, 1000);
RunningJob runningJob = jobClient.submitJob(jobConf);
return runningJob;
}
});
JobID jobId = job.getID();
getTIPId(cluster, jobId);//wait till the map task starts running
return job;
}
// Waits till the map task gets started and gets its tipId from map reports
// and returns the tipId
private TaskID getTIPId(MiniMRCluster cluster,
org.apache.hadoop.mapreduce.JobID jobid) throws Exception {
JobClient client = new JobClient(cluster.createJobConf());
JobID jobId = (JobID) jobid;
TaskReport[] mapReports = null;
TaskID tipId = null;
do { // make sure that the map task is running
Thread.sleep(200);
mapReports = client.getMapTaskReports(jobId);
} while (mapReports.length == 0);
for (TaskReport r : mapReports) {
tipId = r.getTaskID();
break;// because we have only one map
}
JobTracker jt = cluster.getJobTrackerRunner().getJobTracker();
// wait till we see an attempt actually running
while (jt.getTaskStatuses(tipId).length == 0) {
try {
Thread.sleep(100);
} catch(InterruptedException e) {
LOG.warn("Interrupted while waiting for map attempt to be started");
}
}
return tipId;
}
/**
* Make sure that the given user can do killJob using jobdetails.jsp url
* @param cluster
* @param conf
* @param jtURL
* @param jobTrackerJSP
* @param user
* @throws Exception
*/
private void confirmJobDetailsJSPKillJobAsUser(MiniMRCluster cluster,
JobConf conf, String jtURL, String jobTrackerJSP, String user)
throws Exception {
RunningJob job = startSleepJobAsUser(jobSubmitter, conf, cluster);
org.apache.hadoop.mapreduce.JobID jobid = job.getID();
// jobDetailsJSP killJob url
String url = jtURL + "/jobdetails.jsp?" +
"action=kill&jobid="+ jobid.toString();
try {
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(url, user, "POST"));
waitForKillJobToFinish(job);
assertTrue("killJob failed for a job for which user has "
+ "job-modify permission", job.isComplete());
} finally {
if (!job.isComplete()) {
LOG.info("Killing job " + jobid + " from finally block");
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(jobTrackerJSP + "&killJobs=true&jobCheckBox=" +
jobid.toString(), jobSubmitter, "GET"));
}
}
}
static void setupGroupsProvider() throws IOException {
Configuration conf = new Configuration();
conf.set(CommonConfigurationKeys.HADOOP_SECURITY_GROUP_MAPPING,
MyGroupsProvider.class.getName());
Groups.getUserToGroupsMappingService(conf);
MyGroupsProvider.mapping.put(jobSubmitter, Arrays.asList("group1"));
MyGroupsProvider.mapping.put(viewColleague, Arrays.asList("group2"));
MyGroupsProvider.mapping.put(modifyColleague, Arrays.asList("group1"));
MyGroupsProvider.mapping.put(unauthorizedUser, Arrays.asList("evilSociety"));
MyGroupsProvider.mapping.put(mrAdminGroupMember, Arrays.asList(mrAdminGroup));
MyGroupsProvider.mapping.put(viewAndModifyColleague, Arrays.asList("group3"));
MyGroupsProvider.mapping.put(qAdmin, Arrays.asList("group4"));
mrOwner = UserGroupInformation.getCurrentUser().getShortUserName();
MyGroupsProvider.mapping.put(mrOwner, Arrays.asList(
new String[] { "group5", "group6" }));
MyGroupsProvider.mapping.put(jobSubmitter1, Arrays.asList("group7"));
MyGroupsProvider.mapping.put(jobSubmitter2, Arrays.asList("group7"));
MyGroupsProvider.mapping.put(jobSubmitter3, Arrays.asList("group7"));
MyGroupsProvider.mapping.put(mrAdminUser, Arrays.asList("group8"));
}
public void testAuthorizationForJobHistoryPages() throws Exception {
setupGroupsProvider();
Properties props = new Properties();
props.setProperty("hadoop.http.filter.initializers",
DummyFilterInitializer.class.getName());
props.setProperty(JobConf.MR_ACLS_ENABLED, String.valueOf(true));
props.setProperty(QueueManager.toFullPropertyName(
"default", QueueACL.ADMINISTER_JOBS.getAclName()), qAdmin);
props.setProperty(QueueManager.toFullPropertyName(
"default", QueueACL.SUBMIT_JOB.getAclName()), jobSubmitter);
props.setProperty("dfs.permissions", "false");
// Let us have history files on HDFS
props.setProperty("mapred.job.tracker.history.completed.location",
"historyDoneFolderOnHDFS");
props.setProperty(JobConf.MR_ADMINS, mrAdminUser + " " + mrAdminGroup);
startCluster(true, props);
MiniMRCluster cluster = getMRCluster();
int infoPort = cluster.getJobTrackerRunner().getJobTrackerInfoPort();
JobConf conf = new JobConf(cluster.createJobConf());
conf.set(JobContext.JOB_ACL_VIEW_JOB, viewColleague + " group3");
// Let us add group1 and group3 to modify-job-acl. So modifyColleague and
// viewAndModifyColleague will be able to modify the job
conf.set(JobContext.JOB_ACL_MODIFY_JOB, " group1,group3");
final SleepJob sleepJob = new SleepJob();
final JobConf jobConf = new JobConf(conf);
sleepJob.setConf(jobConf);
UserGroupInformation jobSubmitterUGI =
UserGroupInformation.createRemoteUser(jobSubmitter);
RunningJob job =
jobSubmitterUGI.doAs(new PrivilegedExceptionAction<RunningJob>() {
public RunningJob run() throws Exception {
JobClient jobClient = new JobClient(jobConf);
SleepJob sleepJob = new SleepJob();
sleepJob.setConf(jobConf);
JobConf jobConf =
sleepJob.setupJobConf(1, 0, 1000, 1, 1000, 1000);
RunningJob runningJob = jobClient.runJob(jobConf);
return runningJob;
}
});
JobID jobid = job.getID();
JobTracker jobTracker = getMRCluster().getJobTrackerRunner().getJobTracker();
JobInProgress jip = jobTracker.getJob(jobid);
JobConf finalJobConf = jip.getJobConf();
Path doneDir = JobHistory.getCompletedJobHistoryLocation();
// Wait for history file to be written
while(TestJobHistory.getDoneFile(finalJobConf, jobid, doneDir) == null) {
Thread.sleep(2000);
}
Path historyFilePath = new Path(doneDir,
JobHistory.JobInfo.getDoneJobHistoryFileName(finalJobConf, jobid));
String urlEncodedHistoryFileName = URLEncoder.encode(historyFilePath.toString());
String jtURL = "http://localhost:" + infoPort;
// validate access of jobdetails_history.jsp
String jobDetailsJSP =
jtURL + "/jobdetailshistory.jsp?logFile=" + urlEncodedHistoryFileName;
validateViewJob(jobDetailsJSP, "GET");
// validate accesses of jobtaskshistory.jsp
String jobTasksJSP =
jtURL + "/jobtaskshistory.jsp?logFile=" + urlEncodedHistoryFileName;
String[] taskTypes =
new String[] { "JOb_SETUP", "MAP", "REDUCE", "JOB_CLEANUP" };
String[] states =
new String[] { "all", "SUCCEEDED", "FAILED", "KILLED" };
for (String taskType : taskTypes) {
for (String state : states) {
validateViewJob(jobTasksJSP + "&taskType=" + taskType + "&status="
+ state, "GET");
}
}
JobHistory.JobInfo jobInfo = new JobHistory.JobInfo(jobid.toString());
DefaultJobHistoryParser.JobTasksParseListener l =
new DefaultJobHistoryParser.JobTasksParseListener(jobInfo);
JobHistory.parseHistoryFromFS(historyFilePath.toString().substring(5),
l, historyFilePath.getFileSystem(conf));
Map<String, org.apache.hadoop.mapred.JobHistory.Task> tipsMap =
jobInfo.getAllTasks();
for (String tip : tipsMap.keySet()) {
// validate access of taskdetailshistory.jsp
validateViewJob(jtURL + "/taskdetailshistory.jsp?logFile="
+ urlEncodedHistoryFileName + "&tipid=" + tip.toString(), "GET");
Map<String, TaskAttempt> attemptsMap =
tipsMap.get(tip).getTaskAttempts();
for (String attempt : attemptsMap.keySet()) {
// validate access to taskstatshistory.jsp
validateViewJob(jtURL + "/taskstatshistory.jsp?attemptid="
+ attempt.toString() + "&logFile=" + urlEncodedHistoryFileName, "GET");
// validate access to tasklogs
String taskLogURL = TaskLogServlet.getTaskLogUrl("localhost",
attemptsMap.get(attempt).get(Keys.HTTP_PORT), attempt.toString());
validateViewJob(taskLogURL, "GET");
}
}
// For each tip, let us test the effect of deletion of job-acls.xml file and
// deletion of task log dir for each of the attempts of the tip.
// delete job-acls.xml file from the job userlog dir and verify
// if unauthorized users can view task logs of each attempt.
Path jobACLsFilePath = new Path(TaskLog.getJobDir(jobid).toString(),
TaskTracker.jobACLsFile);
new File(jobACLsFilePath.toUri().getPath()).delete();
for (String tip : tipsMap.keySet()) {
Map<String, TaskAttempt> attemptsMap =
tipsMap.get(tip).getTaskAttempts();
for (String attempt : attemptsMap.keySet()) {
String taskLogURL = TaskLogServlet.getTaskLogUrl("localhost",
attemptsMap.get(attempt).get(Keys.HTTP_PORT), attempt.toString());
// unauthorized users can view task logs of each attempt because
// job-acls.xml file is deleted.
assertEquals("Incorrect return code for " + unauthorizedUser,
HttpURLConnection.HTTP_OK, getHttpStatusCode(taskLogURL,
unauthorizedUser, "GET"));
// delete the whole task log dir of attempt and verify that we get
// correct response code (i.e. HTTP_GONE) when task logs are accessed.
File attemptLogDir = TaskLog.getAttemptDir(TaskAttemptID
.forName(attempt), false);
FileUtil.fullyDelete(attemptLogDir);
assertEquals("Incorrect return code for " + jobSubmitter,
HttpURLConnection.HTTP_GONE, getHttpStatusCode(taskLogURL,
jobSubmitter, "GET"));
}
}
// validate access to analysejobhistory.jsp
String analyseJobHistoryJSP =
jtURL + "/analysejobhistory.jsp?logFile=" + urlEncodedHistoryFileName;
validateViewJob(analyseJobHistoryJSP, "GET");
// validate access of jobconf_history.jsp
String jobConfJSP =
jtURL + "/jobconf_history.jsp?logFile=" + urlEncodedHistoryFileName;
validateViewJob(jobConfJSP, "GET");
}
/**
* Starts a sleep job and tries to kill the job using jobdetails.jsp as
* (1) viewColleague (2) unauthorizedUser (3) modifyColleague
* (4) viewAndModifyColleague (5) mrOwner (6) mrAdmin and
* (7) jobSubmitter
*
* Validates the given jsp/servlet against different user names who
* can(or cannot) do both view and modify on the job.
* (1) jobSubmitter, mrOwner and mrAdmin can do both view and modify
* on the job. But we are not validating this in this method. Let the
* caller explicitly validate this, if needed.
* (2) user mentioned in job-view-acls and job-modify-acls can do this
* (3) user mentioned in job-view-acls but not in job-modify-acls cannot
* do this
* (4) user mentioned in job-modify-acls but not in job-view-acls cannot
* do this
* (5) qAdmin cannot do this because he doesn't have view access to the job
* (6) other unauthorized users cannot do this
*
* @throws Exception
*/
private void validateJobDetailsJSPKillJob(MiniMRCluster cluster,
JobConf clusterConf, String jtURL) throws Exception {
JobConf conf = new JobConf(cluster.createJobConf());
conf.set(JobContext.JOB_ACL_VIEW_JOB, viewColleague + " group3");
// Let us add group1 and group3 to modify-job-acl. So modifyColleague and
// viewAndModifyColleague will be able to modify the job
conf.set(JobContext.JOB_ACL_MODIFY_JOB, " group1,group3");
String jobTrackerJSP = jtURL + "/jobtracker.jsp?a=b";
RunningJob job = startSleepJobAsUser(jobSubmitter, conf, cluster);
org.apache.hadoop.mapreduce.JobID jobid = job.getID();
// jobDetailsJSPKillJobAction url
String url = jtURL + "/jobdetails.jsp?" +
"action=kill&jobid="+ jobid.toString();
try {
assertEquals(HttpURLConnection.HTTP_UNAUTHORIZED,
getHttpStatusCode(url, viewColleague, "POST"));
assertEquals(HttpURLConnection.HTTP_UNAUTHORIZED,
getHttpStatusCode(url, unauthorizedUser, "POST"));
assertEquals(HttpURLConnection.HTTP_UNAUTHORIZED,
getHttpStatusCode(url, modifyColleague, "POST"));
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(url, viewAndModifyColleague, "POST"));
waitForKillJobToFinish(job);
assertTrue("killJob using jobdetails.jsp failed for a job for which "
+ "user has job-view and job-modify permissions", job.isComplete());
} finally {
if (!job.isComplete()) {
LOG.info("Killing job " + jobid + " from finally block");
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(jobTrackerJSP + "&killJobs=true&jobCheckBox=" +
jobid.toString(), jobSubmitter, "GET"));
}
}
// check if jobSubmitter, mrOwner and mrAdmin can do
// killJob using jobdetails.jsp url
confirmJobDetailsJSPKillJobAsUser(cluster, conf, jtURL, jobTrackerJSP,
jobSubmitter);
confirmJobDetailsJSPKillJobAsUser(cluster, conf, jtURL, jobTrackerJSP,
mrOwner);
confirmJobDetailsJSPKillJobAsUser(cluster, conf, jtURL, jobTrackerJSP,
mrAdminUser);
confirmJobDetailsJSPKillJobAsUser(cluster, conf, jtURL, jobTrackerJSP,
mrAdminGroupMember);
confirmJobDetailsJSPKillJobAsUser(cluster, conf, jtURL, jobTrackerJSP,
qAdmin);
}
/**
* Make sure that the given user can do killJob using jobtracker.jsp url
* @param cluster
* @param conf
* @param jtURL
* @param user
* @throws Exception
*/
private void confirmJobTrackerJSPKillJobAsUser(MiniMRCluster cluster,
JobConf conf, String jtURL, String user)
throws Exception {
String jobTrackerJSP = jtURL + "/jobtracker.jsp?a=b";
RunningJob job = startSleepJobAsUser(jobSubmitter, conf, cluster);
org.apache.hadoop.mapreduce.JobID jobid = job.getID();
// jobTrackerJSP killJob url
String url = jobTrackerJSP +
"&killJobs=true&jobCheckBox=" + jobid.toString();
try {
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(url, user, "POST"));
waitForKillJobToFinish(job);
assertTrue("killJob failed for a job for which user has "
+ "job-modify permission", job.isComplete());
} finally {
if (!job.isComplete()) {
LOG.info("Killing job " + jobid + " from finally block");
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(jobTrackerJSP + "&killJobs=true&jobCheckBox=" +
jobid.toString(), jobSubmitter, "GET"));
}
}
}
/**
* Waits for the job to become completed as we know that killJob is called
* @param job
* @throws IOException
*/
void waitForKillJobToFinish(RunningJob job) throws IOException {
while (!job.isComplete()) {
try {
Thread.sleep(200);
} catch (InterruptedException e){
LOG.warn("Interrupted while waiting for killJob() to finish for "
+ job.getID());
}
}
}
/**
* Make sure that multiple jobs get killed using jobtracker.jsp url when
* user has modify access on only some of those jobs.
* @param cluster
* @param conf
* @param jtURL
* @param user
* @throws Exception
*/
private void validateKillMultipleJobs(MiniMRCluster cluster,
JobConf conf, String jtURL) throws Exception {
String jobTrackerJSP = jtURL + "/jobtracker.jsp?a=b";
// jobTrackerJSP killJob url
String url = jobTrackerJSP + "&killJobs=true";
// view-job-acl doesn't matter for killJob from jobtracker jsp page
conf.set(JobContext.JOB_ACL_VIEW_JOB, " ");
// Let us start 4 jobs as 4 different users(none of these 4 users is
// mrOwner and none of these users is a member of mrAdmin and none of
// these 4 users is a queue admin for the default queue). So only
// based on the config JobContext.JOB_ACL_MODIFY_JOB being set here and the
// job-submitter, killJob on each of the jobs will be succeeded.
// start 1st job.
// Out of these 4 users, only jobSubmitter can do killJob on 1st job
conf.set(JobContext.JOB_ACL_MODIFY_JOB, " ");
RunningJob job1 = startSleepJobAsUser(jobSubmitter, conf, cluster);
org.apache.hadoop.mapreduce.JobID jobid = job1.getID();
url = url.concat("&jobCheckBox=" + jobid.toString());
// start 2nd job.
// Out of these 4 users, only jobSubmitter1 can do killJob on 2nd job
RunningJob job2 = startSleepJobAsUser(jobSubmitter1, conf, cluster);
jobid = job2.getID();
url = url.concat("&jobCheckBox=" + jobid.toString());
// start 3rd job.
// Out of these 4 users, only jobSubmitter2 can do killJob on 3rd job
RunningJob job3 = startSleepJobAsUser(jobSubmitter2, conf, cluster);
jobid = job3.getID();
url = url.concat("&jobCheckBox=" + jobid.toString());
// start 4rd job.
// Out of these 4 users, jobSubmitter1 and jobSubmitter3
// can do killJob on 4th job
conf.set(JobContext.JOB_ACL_MODIFY_JOB, jobSubmitter1);
RunningJob job4 = startSleepJobAsUser(jobSubmitter3,
conf, cluster);
jobid = job4.getID();
url = url.concat("&jobCheckBox=" + jobid.toString());
try {
// Try killing all the 4 jobs as user viewColleague who can kill only
// 2nd and 4th jobs. Check if 1st and 3rd jobs are not killed and
// 2nd and 4th jobs get killed
assertEquals(HttpURLConnection.HTTP_UNAUTHORIZED,
getHttpStatusCode(url, jobSubmitter1, "POST"));
waitForKillJobToFinish(job2);
assertTrue("killJob failed for a job for which user has "
+ "job-modify permission", job2.isComplete());
waitForKillJobToFinish(job4);
assertTrue("killJob failed for a job for which user has "
+ "job-modify permission", job4.isComplete());
assertFalse("killJob succeeded for a job for which user doesnot "
+ " have job-modify permission", job1.isComplete());
assertFalse("killJob succeeded for a job for which user doesnot "
+ " have job-modify permission", job3.isComplete());
} finally {
// Kill all 4 jobs as user mrOwner(even though some of them
// were already killed)
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(url, mrOwner, "GET"));
}
}
/**
* Run a job and validate if JSPs/Servlets are going through authentication
* and authorization.
* @throws Exception
*/
@Test
public void testWebUIAuthorization() throws Exception {
setupGroupsProvider();
Properties props = new Properties();
props.setProperty("hadoop.http.filter.initializers",
DummyFilterInitializer.class.getName());
props.setProperty(JobConf.MR_ACLS_ENABLED, String.valueOf(true));
props.setProperty(QueueManager.toFullPropertyName(
"default", QueueACL.ADMINISTER_JOBS.getAclName()), qAdmin);
props.setProperty(QueueManager.toFullPropertyName(
"default", QueueACL.SUBMIT_JOB.getAclName()),
jobSubmitter + "," + jobSubmitter1 + "," + jobSubmitter2 + "," +
jobSubmitter3);
props.setProperty("dfs.permissions", "false");
// let us have enough map slots so that there won't be waiting for slots
props.setProperty("mapred.tasktracker.map.tasks.maximum", "6");
props.setProperty(JSPUtil.PRIVATE_ACTIONS_KEY, "true");
props.setProperty(JobConf.MR_ADMINS, mrAdminUser + " " + mrAdminGroup);
startCluster(true, props);
MiniMRCluster cluster = getMRCluster();
int infoPort = cluster.getJobTrackerRunner().getJobTrackerInfoPort();
JobConf clusterConf = cluster.createJobConf();
JobConf conf = new JobConf(clusterConf);
conf.set(JobContext.JOB_ACL_VIEW_JOB, viewColleague + " group3");
// Let us add group1 and group3 to modify-job-acl. So modifyColleague and
// viewAndModifyColleague will be able to modify the job
conf.set(JobContext.JOB_ACL_MODIFY_JOB, " group1,group3");
RunningJob job = startSleepJobAsUser(jobSubmitter, conf, cluster);
org.apache.hadoop.mapreduce.JobID jobid = job.getID();
String jtURL = "http://localhost:" + infoPort;
String jobTrackerJSP = jtURL + "/jobtracker.jsp?a=b";
try {
// Currently, authorization is not done for jobtracker page. So allow
// everyone to view it.
validateJobTrackerJSPAccess(jtURL);
validateJobDetailsJSPAccess(jobid, jtURL);
validateTaskGraphServletAccess(jobid, jtURL);
validateJobTasksJSPAccess(jobid, jtURL);
validateJobConfJSPAccess(jobid, jtURL);
validateJobFailuresJSPAccess(jobid, jtURL);
valiateJobBlacklistedTrackerJSPAccess(jobid, jtURL);
validateJobTrackerJSPSetPriorityAction(jobid, jtURL);
// Wait for the tip to start so as to test task related JSP
TaskID tipId = getTIPId(cluster, jobid);
validateTaskStatsJSPAccess(jobid, jtURL, tipId);
validateTaskDetailsJSPAccess(jobid, jtURL, tipId);
validateJobTrackerJSPKillJobAction(job, jtURL);
} finally {
if (!job.isComplete()) { // kill the job(as jobSubmitter) if needed
LOG.info("Killing job " + jobid + " from finally block");
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(jobTrackerJSP + "&killJobs=true&jobCheckBox=" +
jobid.toString(), jobSubmitter, "GET"));
}
}
// validate killJob of jobdetails.jsp
validateJobDetailsJSPKillJob(cluster, clusterConf, jtURL);
// validate killJob of jobtracker.jsp as users viewAndModifyColleague,
// jobSubmitter, mrOwner and mrAdmin
confirmJobTrackerJSPKillJobAsUser(cluster, conf, jtURL,
viewAndModifyColleague);
confirmJobTrackerJSPKillJobAsUser(cluster, conf, jtURL, jobSubmitter);
confirmJobTrackerJSPKillJobAsUser(cluster, conf, jtURL, mrOwner);
confirmJobTrackerJSPKillJobAsUser(cluster, conf, jtURL, mrAdminUser);
confirmJobTrackerJSPKillJobAsUser(cluster, conf, jtURL, mrAdminGroupMember);
confirmJobTrackerJSPKillJobAsUser(cluster, conf, jtURL, qAdmin);
// validate killing of multiple jobs using jobtracker jsp and check
// if all the jobs which can be killed by user are actually the ones that
// got killed
validateKillMultipleJobs(cluster, conf, jtURL);
}
public void testWebUIAuthorizationForCommonServlets() throws Exception {
setupGroupsProvider();
Properties props = new Properties();
props.setProperty("hadoop.http.filter.initializers",
DummyFilterInitializer.class.getName());
props.setProperty(CommonConfigurationKeys.HADOOP_SECURITY_AUTHORIZATION, "true");
props.setProperty(JobConf.MR_ADMINS, mrAdminUser + " " + mrAdminGroup);
startCluster(true, props);
validateCommonServlets(getMRCluster());
}
private void validateCommonServlets(MiniMRCluster cluster) throws IOException {
int infoPort = cluster.getJobTrackerRunner().getJobTrackerInfoPort();
String jtURL = "http://localhost:" + infoPort;
for (String servlet : new String[] { "logs", "stacks", "logLevel" }) {
String url = jtURL + "/" + servlet;
checkAccessToCommonServlet(url);
}
// validate access to common servlets for TaskTracker.
String ttURL = "http://localhost:"
+ cluster.getTaskTrackerRunner(0).getTaskTracker().getHttpPort();
for (String servlet : new String[] { "logs", "stacks", "logLevel" }) {
String url = ttURL + "/" + servlet;
checkAccessToCommonServlet(url);
}
}
private void checkAccessToCommonServlet(String url) throws IOException {
url = url + "?a=b";
assertEquals(HttpURLConnection.HTTP_OK, getHttpStatusCode(url, mrAdminUser,
"GET"));
assertEquals(HttpURLConnection.HTTP_OK, getHttpStatusCode(url,
mrAdminGroupMember, "GET"));
assertEquals(HttpURLConnection.HTTP_OK, getHttpStatusCode(url,
mrOwner, "GET"));
// no access for any other user
assertEquals(HttpURLConnection.HTTP_UNAUTHORIZED, getHttpStatusCode(url,
jobSubmitter, "GET"));
}
// validate killJob of jobtracker.jsp
private void validateJobTrackerJSPKillJobAction(RunningJob job, String jtURL)
throws IOException {
org.apache.hadoop.mapreduce.JobID jobid = job.getID();
String jobTrackerJSP = jtURL + "/jobtracker.jsp?a=b";
String jobTrackerJSPKillJobAction = jobTrackerJSP +
"&killJobs=true&jobCheckBox="+ jobid.toString();
validateModifyJob(jobTrackerJSPKillJobAction, "GET");
waitForKillJobToFinish(job);
assertTrue("killJob failed for a job for which user has "
+ "job-modify permission", job.isComplete());
}
// validate viewing of job of taskdetails.jsp
private void validateTaskDetailsJSPAccess(
org.apache.hadoop.mapreduce.JobID jobid, String jtURL, TaskID tipId)
throws IOException {
String taskDetailsJSP = jtURL + "/taskdetails.jsp?jobid=" +
jobid.toString() + "&tipid=" + tipId;
validateViewJob(taskDetailsJSP, "GET");
}
// validate taskstats.jsp
private void validateTaskStatsJSPAccess(
org.apache.hadoop.mapreduce.JobID jobid, String jtURL, TaskID tipId)
throws IOException {
String taskStatsJSP = jtURL + "/taskstats.jsp?jobid=" +
jobid.toString() + "&tipid=" + tipId;
validateViewJob(taskStatsJSP, "GET");
}
// validate setJobPriority
private void validateJobTrackerJSPSetPriorityAction(
org.apache.hadoop.mapreduce.JobID jobid, String jtURL)
throws IOException {
String jobTrackerJSP = jtURL + "/jobtracker.jsp?a=b";
String jobTrackerJSPSetJobPriorityAction = jobTrackerJSP +
"&changeJobPriority=true&setJobPriority="+"HIGH"+"&jobCheckBox=" +
jobid.toString();
validateModifyJob(jobTrackerJSPSetJobPriorityAction, "GET");
// jobSubmitter, mrOwner, qAdmin and mrAdmin are not validated for
// job-modify permission in validateModifyJob(). So let us do it
// explicitly here
assertEquals(HttpURLConnection.HTTP_OK, getHttpStatusCode(
jobTrackerJSPSetJobPriorityAction, jobSubmitter, "GET"));
assertEquals(HttpURLConnection.HTTP_OK, getHttpStatusCode(
jobTrackerJSPSetJobPriorityAction, mrAdminUser, "GET"));
assertEquals(HttpURLConnection.HTTP_OK, getHttpStatusCode(
jobTrackerJSPSetJobPriorityAction, mrAdminGroupMember, "GET"));
assertEquals(HttpURLConnection.HTTP_OK, getHttpStatusCode(
jobTrackerJSPSetJobPriorityAction, qAdmin, "GET"));
assertEquals(HttpURLConnection.HTTP_OK, getHttpStatusCode(
jobTrackerJSPSetJobPriorityAction, mrOwner, "GET"));
}
// validate access of jobblacklistedtrackers.jsp
private void valiateJobBlacklistedTrackerJSPAccess(
org.apache.hadoop.mapreduce.JobID jobid, String jtURL)
throws IOException {
String jobBlacklistedTrackersJSP = jtURL +
"/jobblacklistedtrackers.jsp?jobid="+jobid.toString();
validateViewJob(jobBlacklistedTrackersJSP, "GET");
}
// validate access of jobfailures.jsp
private void validateJobFailuresJSPAccess(
org.apache.hadoop.mapreduce.JobID jobid, String jtURL)
throws IOException {
String jobFailuresJSP = jtURL + "/jobfailures.jsp?jobid="+jobid.toString();
validateViewJob(jobFailuresJSP, "GET");
}
// validate access of jobconf.jsp
private void validateJobConfJSPAccess(
org.apache.hadoop.mapreduce.JobID jobid, String jtURL)
throws IOException {
String jobConfJSP = jtURL + "/jobconf.jsp?jobid="+jobid.toString();
validateViewJob(jobConfJSP, "GET");
}
// validate access of jobtasks.jsp
private void validateJobTasksJSPAccess(
org.apache.hadoop.mapreduce.JobID jobid, String jtURL)
throws IOException {
String jobTasksJSP = jtURL + "/jobtasks.jsp?jobid="
+ jobid.toString() + "&type=map&pagenum=1&state=running";
validateViewJob(jobTasksJSP, "GET");
}
// validate access of TaskGraphServlet
private void validateTaskGraphServletAccess(
org.apache.hadoop.mapreduce.JobID jobid, String jtURL)
throws IOException {
String taskGraphServlet = jtURL + "/taskgraph?type=map&jobid="
+ jobid.toString();
validateViewJob(taskGraphServlet, "GET");
taskGraphServlet = jtURL + "/taskgraph?type=reduce&jobid="
+ jobid.toString();
validateViewJob(taskGraphServlet, "GET");
}
// validate access of jobdetails.jsp
private void validateJobDetailsJSPAccess(
org.apache.hadoop.mapreduce.JobID jobid, String jtURL)
throws IOException {
String jobDetailsJSP = jtURL + "/jobdetails.jsp?jobid="
+ jobid.toString();
validateViewJob(jobDetailsJSP, "GET");
}
// validate access of jobtracker.jsp
private void validateJobTrackerJSPAccess(String jtURL)
throws IOException {
String jobTrackerJSP = jtURL + "/jobtracker.jsp?a=b";
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(jobTrackerJSP, jobSubmitter, "GET"));
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(jobTrackerJSP, viewColleague, "GET"));
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(jobTrackerJSP, unauthorizedUser, "GET"));
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(jobTrackerJSP, modifyColleague, "GET"));
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(jobTrackerJSP, viewAndModifyColleague, "GET"));
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(jobTrackerJSP, mrOwner, "GET"));
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(jobTrackerJSP, qAdmin, "GET"));
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(jobTrackerJSP, mrAdminUser, "GET"));
assertEquals(HttpURLConnection.HTTP_OK,
getHttpStatusCode(jobTrackerJSP, mrAdminGroupMember, "GET"));
}
}