[
https://issues.apache.org/jira/browse/YARN-11674?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17840843#comment-17840843
]
ASF GitHub Bot commented on YARN-11674:
---------------------------------------
tomicooler commented on code in PR #6751:
URL: https://github.com/apache/hadoop/pull/6751#discussion_r1579625169
##########
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/linux/resources/CGroupsV2CpuResourceHandlerImpl.java:
##########
@@ -0,0 +1,99 @@
+/**
+ * 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
+ * <p/>
+ * http://www.apache.org/licenses/LICENSE-2.0
+ * <p/>
+ * 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 org.apache.hadoop.classification.InterfaceAudience;
+import org.apache.hadoop.classification.InterfaceStability;
+import org.apache.hadoop.classification.VisibleForTesting;
+
+/**
+ * An implementation for using CGroups V2 to restrict CPU usage on Linux. The
+ * implementation supports 3 different controls - restrict usage of all YARN
+ * containers, restrict relative usage of individual YARN containers and
+ * restrict usage of individual YARN containers. Admins can set the overall CPU
+ * to be used by all YARN containers - this is implemented by setting
+ * cpu.max to the value desired. If strict resource usage mode is not enabled,
+ * cpu.weight is set for individual containers - this prevents containers from
+ * exceeding the overall limit for YARN containers but individual containers
+ * can use as much of the CPU as available(under the YARN limit). If strict
+ * resource usage is enabled, then container can only use the percentage of
+ * CPU allocated to them and this is again implemented using cpu.max.
+ */
+@InterfaceStability.Unstable
+@InterfaceAudience.Private
+public class CGroupsV2CpuResourceHandlerImpl extends
AbstractCGroupsCpuResourceHandler {
+ private static final CGroupsHandler.CGroupController CPU =
+ CGroupsHandler.CGroupController.CPU;
+
+ @VisibleForTesting
+ static final int CPU_DEFAULT_WEIGHT = 100; // cgroup v2 default
+ static final int CPU_DEFAULT_WEIGHT_OPPORTUNISTIC = 1;
+ static final int CPU_MAX_WEIGHT = 10000;
+ static final String NO_LIMIT = "max";
+
+
+ CGroupsV2CpuResourceHandlerImpl(CGroupsHandler cGroupsHandler) {
+ super(cGroupsHandler);
+ }
+
+ @Override
+ protected void updateCgroupMaxCpuLimit(String cgroupId, String max, String
period)
+ throws ResourceHandlerException {
+ String cpuMaxLimit = cGroupsHandler.getCGroupParam(CPU, cgroupId,
Review Comment:
Maybe a small doc here about the file format:
https://docs.kernel.org/admin-guide/cgroup-v2.html#cpu-interface-files
```
A read-write two value file which exists on non-root cgroups. The default is
“max 100000”.
The maximum bandwidth limit. It’s in the following format:
$MAX $PERIOD
which indicates that the group may consume up to $MAX in each $PERIOD
duration. “max” for $MAX indicates no limit. If only one number is written,
$MAX is updated.
```
rename: cpuMaxLimit -> currentCpuMax
##########
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/linux/resources/AbstractCGroupsCpuResourceHandler.java:
##########
@@ -0,0 +1,219 @@
+/**
+ * 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
+ * <p/>
+ * http://www.apache.org/licenses/LICENSE-2.0
+ * <p/>
+ * 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 org.apache.hadoop.classification.InterfaceAudience;
+import org.apache.hadoop.classification.InterfaceStability;
+import org.apache.hadoop.classification.VisibleForTesting;
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.yarn.api.records.ContainerId;
+import org.apache.hadoop.yarn.api.records.ExecutionType;
+import org.apache.hadoop.yarn.api.records.Resource;
+import org.apache.hadoop.yarn.conf.YarnConfiguration;
+import org.apache.hadoop.yarn.security.ContainerTokenIdentifier;
+import
org.apache.hadoop.yarn.server.nodemanager.containermanager.container.Container;
+import
org.apache.hadoop.yarn.server.nodemanager.containermanager.linux.privileged.PrivilegedOperation;
+import org.apache.hadoop.yarn.server.nodemanager.util.NodeManagerHardwareUtils;
+import org.apache.hadoop.yarn.util.ResourceCalculatorPlugin;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.io.File;
+import java.util.ArrayList;
+import java.util.List;
+
+@InterfaceStability.Unstable
+@InterfaceAudience.Private
+public abstract class AbstractCGroupsCpuResourceHandler implements
CpuResourceHandler {
+
+ static final Logger LOG =
+ LoggerFactory.getLogger(AbstractCGroupsCpuResourceHandler.class);
+
+ protected CGroupsHandler cGroupsHandler;
+ private boolean strictResourceUsageMode = false;
+ private float yarnProcessors;
+ private int nodeVCores;
+ private static final CGroupsHandler.CGroupController CPU =
+ CGroupsHandler.CGroupController.CPU;
+
+ @VisibleForTesting
+ static final int MAX_QUOTA_US = 1000 * 1000;
+ @VisibleForTesting
+ static final int MIN_PERIOD_US = 1000;
+
+ AbstractCGroupsCpuResourceHandler(CGroupsHandler cGroupsHandler) {
+ this.cGroupsHandler = cGroupsHandler;
+ }
+
+ @Override
+ public List<PrivilegedOperation> bootstrap(Configuration conf)
+ throws ResourceHandlerException {
+ return bootstrap(
+ ResourceCalculatorPlugin.getResourceCalculatorPlugin(null, conf),
conf);
+ }
+
+ @VisibleForTesting
+ List<PrivilegedOperation> bootstrap(
+ ResourceCalculatorPlugin plugin, Configuration conf)
+ throws ResourceHandlerException {
+ this.strictResourceUsageMode = conf.getBoolean(
+ YarnConfiguration.NM_LINUX_CONTAINER_CGROUPS_STRICT_RESOURCE_USAGE,
+
YarnConfiguration.DEFAULT_NM_LINUX_CONTAINER_CGROUPS_STRICT_RESOURCE_USAGE);
+ this.cGroupsHandler.initializeCGroupController(CPU);
+ nodeVCores = NodeManagerHardwareUtils.getVCores(plugin, conf);
+
+ // cap overall usage to the number of cores allocated to YARN
+ yarnProcessors = NodeManagerHardwareUtils.getContainersCPUs(plugin, conf);
+ int systemProcessors = NodeManagerHardwareUtils.getNodeCPUs(plugin, conf);
+ boolean existingCpuLimits;
+ existingCpuLimits = cpuLimitExists(
+ cGroupsHandler.getPathForCGroup(CPU, ""));
+
+ if (systemProcessors != (int) yarnProcessors) {
+ LOG.info("YARN containers restricted to " + yarnProcessors + " cores");
+ int[] limits = getOverallLimits(yarnProcessors);
+ updateCgroupMaxCpuLimit("", String.valueOf(limits[1]),
String.valueOf(limits[0]));
+ } else if (existingCpuLimits) {
+ LOG.info("Removing CPU constraints for YARN containers.");
+ updateCgroupMaxCpuLimit("", String.valueOf(-1), null);
+ }
+ return null;
+ }
+
+ protected abstract void updateCgroupMaxCpuLimit(String cgroupId, String
quota, String period)
+ throws ResourceHandlerException;
+ protected abstract boolean cpuLimitExists(String path) throws
ResourceHandlerException;
+
+
+ @VisibleForTesting
+ @InterfaceAudience.Private
+ public static int[] getOverallLimits(float yarnProcessors) {
Review Comment:
NIT. For readability we could introduce a wrapper over this that return an
object with explicit `period` and `quota` fields. The newly introduced
`updateCgroupMaxCpuLimit` and alike methods could use this object as parameter.
Since `getOverallLimits` is used in the depcreated LCE code as well we, must
keep this too, so I'm fine by keeping it as is.
> Update CpuResourceHandler implementation for cgroup v2 support
> --------------------------------------------------------------
>
> Key: YARN-11674
> URL: https://issues.apache.org/jira/browse/YARN-11674
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Benjamin Teke
> Assignee: Benjamin Teke
> Priority: Major
> Labels: pull-request-available
>
> cgroup v2 has some changes in various controllers (some changed their
> functionality, some were removed). This task is about checking if
> CpuResourceHandler's
> [implementation|https://github.com/apache/hadoop/blob/d336227e5c63a70db06ac26697994c96ed89d230/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/linux/resources/CGroupsCpuResourceHandlerImpl.java#L60]
> need any updates.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
