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

HDDS-921. Add JVM pause monitor to Ozone Daemons (OM, SCM and Datanodes). Contributed by Bharat Viswanadham.

This commit is contained in:
Bharat Viswanadham 2018-12-23 17:33:16 -08:00
parent d944d5ec46
commit 26e4be7022
3 changed files with 39 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/statemachine/DatanodeStateMachine.java
View File

@ -49,6 +49,7 @@ import org.apache.hadoop.ozone.container.replication.DownloadAndImportReplicator
import org.apache.hadoop.ozone.container.replication.ReplicationSupervisor;
import org.apache.hadoop.ozone.container.replication.SimpleContainerDownloader;
import org.apache.hadoop.ozone.protocol.commands.SCMCommand;
import org.apache.hadoop.util.JvmPauseMonitor;
import org.apache.hadoop.util.Time;
import org.apache.hadoop.util.concurrent.HadoopExecutors;
@ -78,6 +79,8 @@ public class DatanodeStateMachine implements Closeable {
private Thread cmdProcessThread = null;
private final ReplicationSupervisor supervisor;
private JvmPauseMonitor jvmPauseMonitor;
/**
* Constructs a a datanode state machine.
*
@ -159,6 +162,12 @@ public class DatanodeStateMachine implements Closeable {
container.start();
reportManager.init();
initCommandHandlerThread(conf);
// Start jvm monitor
jvmPauseMonitor = new JvmPauseMonitor();
jvmPauseMonitor.init(conf);
jvmPauseMonitor.start();
while (context.getState() != DatanodeStates.SHUTDOWN) {
try {
LOG.debug("Executing cycle Number : {}", context.getExecutionCount());
@ -242,6 +251,10 @@ public class DatanodeStateMachine implements Closeable {
if(container != null) {
container.stop();
}
if (jvmPauseMonitor != null) {
jvmPauseMonitor.stop();
}
}
/**

16
hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/StorageContainerManager.java
View File

@ -84,6 +84,7 @@ import org.apache.hadoop.ozone.common.StorageInfo;
import org.apache.hadoop.ozone.lease.LeaseManager;
import org.apache.hadoop.security.UserGroupInformation;
import org.apache.hadoop.util.GenericOptionsParser;
import org.apache.hadoop.util.JvmPauseMonitor;
import org.apache.hadoop.util.StringUtils;
import static org.apache.hadoop.hdds.scm.ScmConfigKeys
@ -183,6 +184,9 @@ public final class StorageContainerManager extends ServiceRuntimeInfoImpl
private final ReplicationActivityStatus replicationStatus;
private final SCMChillModeManager scmChillModeManager;
private JvmPauseMonitor jvmPauseMonitor;
private final OzoneConfiguration configuration;
/**
* Creates a new StorageContainerManager. Configuration will be updated
* with information on the
@ -192,6 +196,7 @@ public final class StorageContainerManager extends ServiceRuntimeInfoImpl
*/
private StorageContainerManager(OzoneConfiguration conf) throws IOException {
configuration = conf;
StorageContainerManager.initMetrics();
initContainerReportCache(conf);
@ -308,6 +313,7 @@ public final class StorageContainerManager extends ServiceRuntimeInfoImpl
eventQueue.addHandler(SCMEvents.NODE_REGISTRATION_CONT_REPORT,
scmChillModeManager);
registerMXBean();
}
/**
@ -683,6 +689,12 @@ public final class StorageContainerManager extends ServiceRuntimeInfoImpl
scmBlockManager.start();
replicationStatus.start();
replicationManager.start();
// Start jvm monitor
jvmPauseMonitor = new JvmPauseMonitor();
jvmPauseMonitor.init(configuration);
jvmPauseMonitor.start();
setStartTime();
}
@ -766,6 +778,10 @@ public final class StorageContainerManager extends ServiceRuntimeInfoImpl
} catch (Exception ex) {
LOG.error("SCM Event Queue stop failed", ex);
}
if (jvmPauseMonitor != null) {
jvmPauseMonitor.stop();
}
IOUtils.cleanupWithLogger(LOG, containerManager);
IOUtils.cleanupWithLogger(LOG, pipelineManager);
}

10
hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/OzoneManager.java
View File

@ -85,6 +85,7 @@ import org.apache.hadoop.ozone.security.acl.OzoneObjInfo;
import org.apache.hadoop.ozone.security.acl.RequestContext;
import org.apache.hadoop.security.UserGroupInformation;
import org.apache.hadoop.util.GenericOptionsParser;
import org.apache.hadoop.util.JvmPauseMonitor;
import org.apache.hadoop.util.ReflectionUtils;
import org.apache.hadoop.util.ShutdownHookManager;
import org.apache.hadoop.util.StringUtils;
@ -178,6 +179,7 @@ public final class OzoneManager extends ServiceRuntimeInfoImpl
private final File omMetaDir;
private final boolean isAclEnabled;
private final IAccessAuthorizer accessAuthorizer;
private JvmPauseMonitor jvmPauseMonitor;
private OzoneManager(OzoneConfiguration conf) throws IOException {
Preconditions.checkNotNull(conf);
@ -639,6 +641,11 @@ public final class OzoneManager extends ServiceRuntimeInfoImpl
httpServer = new OzoneManagerHttpServer(configuration, this);
httpServer.start();
registerMXBean();
// Start jvm monitor
jvmPauseMonitor = new JvmPauseMonitor();
jvmPauseMonitor.init(configuration);
jvmPauseMonitor.start();
setStartTime();
}
@ -660,6 +667,9 @@ public final class OzoneManager extends ServiceRuntimeInfoImpl
metadataManager.stop();
metrics.unRegister();
unregisterMXBean();
if (jvmPauseMonitor != null) {
jvmPauseMonitor.stop();
}
} catch (Exception e) {
LOG.error("OzoneManager stop failed.", e);
}