[ 
https://issues.apache.org/jira/browse/YARN-6668?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16056730#comment-16056730
 ] 

ASF GitHub Bot commented on YARN-6668:
--------------------------------------

Github user haibchen commented on a diff in the pull request:

    https://github.com/apache/hadoop/pull/241#discussion_r123130198
  
    --- Diff: 
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/linux/resources/CGroupsResourceCalculator.java
 ---
    @@ -0,0 +1,346 @@
    +/**
    + * 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.yarn.server.nodemanager.containermanager.linux.resources;
    +
    +import com.google.common.annotations.VisibleForTesting;
    +import org.apache.hadoop.util.CpuTimeTracker;
    +import org.apache.hadoop.util.Shell;
    +import org.apache.hadoop.util.SysInfoLinux;
    +import org.apache.hadoop.yarn.exceptions.YarnException;
    +import org.apache.hadoop.yarn.util.Clock;
    +import org.apache.hadoop.yarn.util.ResourceCalculatorProcessTree;
    +import org.apache.hadoop.yarn.util.SystemClock;
    +
    +import java.io.BufferedReader;
    +import java.io.File;
    +import java.io.FileInputStream;
    +import java.io.FileNotFoundException;
    +import java.io.IOException;
    +import java.io.InputStreamReader;
    +import java.math.BigInteger;
    +import java.nio.charset.Charset;
    +import java.util.function.Function;
    +import java.util.regex.Matcher;
    +import java.util.regex.Pattern;
    +
    +/**
    + * A cgroups file-system based Resource calculator without the process tree
    + * features.
    + */
    +public class CGroupsResourceCalculator extends 
ResourceCalculatorProcessTree {
    +  enum Result {
    +    Continue,
    +    Exit
    +  }
    +  private static final String PROCFS = "/proc";
    +  static final String CGROUP = "cgroup";
    +  static final String CPU_STAT = "cpuacct.stat";
    +  static final String MEM_STAT = "memory.usage_in_bytes";
    +  static final String MEMSW_STAT = "memory.memsw.usage_in_bytes";
    +  private static final String USER = "user ";
    +  private static final String SYSTEM = "system ";
    +
    +  private static final Pattern CGROUP_FILE_FORMAT = Pattern.compile(
    +      "^(\\d+):([^:]+):/(.*)$");
    +  private final String procfsDir;
    +  private CGroupsHandler cGroupsHandler;
    +
    +  private String pid;
    +  private File cpuStat;
    +  private File memStat;
    +  private File memswStat;
    +
    +  private final long jiffyLengthMs;
    +  private BigInteger processTotalJiffies = BigInteger.ZERO;
    +  private final CpuTimeTracker cpuTimeTracker;
    +  private Clock clock;
    +
    +  private final static Object LOCK = new Object();
    +  private static boolean firstError = true;
    +
    +  /**
    +   * Create resource calculator for all Yarn containers.
    +   * @throws YarnException Could not access cgroups
    +   */
    +  public CGroupsResourceCalculator() throws YarnException {
    +    this(null, PROCFS, ResourceHandlerModule.getCGroupsHandler(),
    +        SystemClock.getInstance());
    +  }
    +
    +  /**
    +   * Create resource calculator for the container that has the specified 
pid.
    +   * @param pid A pid from the cgroup or null for all containers
    +   * @throws YarnException Could not access cgroups
    +   */
    +  public CGroupsResourceCalculator(String pid) throws YarnException {
    +    this(pid, PROCFS, ResourceHandlerModule.getCGroupsHandler(),
    +        SystemClock.getInstance());
    +  }
    +
    +  /**
    +   * Create resource calculator for testing.
    +   * @param pid A pid from the cgroup or null for all containers
    +   * @param procfsDir Path to /proc or a mock /proc directory
    +   * @param cGroupsHandler Initialized cgroups handler object
    +   * @param clock A clock object
    +   * @throws YarnException YarnException Could not access cgroups
    +   */
    +  @VisibleForTesting
    +  CGroupsResourceCalculator(String pid, String procfsDir,
    +                            CGroupsHandler cGroupsHandler, Clock clock)
    +      throws YarnException {
    +    super(pid);
    +    this.procfsDir = procfsDir;
    +    this.cGroupsHandler = cGroupsHandler;
    +    this.pid = pid;
    +    this.cpuTimeTracker =
    +        new CpuTimeTracker(SysInfoLinux.JIFFY_LENGTH_IN_MILLIS);
    +    this.clock = clock;
    +    this.jiffyLengthMs = (clock == SystemClock.getInstance()) ?
    +      SysInfoLinux.JIFFY_LENGTH_IN_MILLIS : 10;
    +    setCGroupFilePaths();
    +  }
    +
    +  @Override
    +  public float getCpuUsagePercent() {
    +    readTotalProcessJiffies();
    +    cpuTimeTracker.updateElapsedJiffies(
    +        processTotalJiffies,
    +        clock.getTime());
    +    return cpuTimeTracker.getCpuTrackerUsagePercent();
    +  }
    +
    +  @Override
    +  public long getCumulativeCpuTime() {
    --- End diff --
    
    Why not use cpuacct.usage?


> Use cgroup to get container resource utilization
> ------------------------------------------------
>
>                 Key: YARN-6668
>                 URL: https://issues.apache.org/jira/browse/YARN-6668
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: nodemanager
>    Affects Versions: 3.0.0-alpha3
>            Reporter: Haibo Chen
>            Assignee: Miklos Szegedi
>
> Container Monitor relies on proc file system to get container resource 
> utilization, which is not as efficient as reading cgroup accounting. We 
> should in NM, when cgroup is enabled, read cgroup stats instead. 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org

Reply via email to