This is an automated email from the ASF dual-hosted git repository.
adoroszlai pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/ozone.git
The following commit(s) were added to refs/heads/master by this push:
new 0f56dfd213 HDDS-8731. Standardize JVM pause monitor (#4814)
0f56dfd213 is described below
commit 0f56dfd2135cb45b63a3a69398e6e7ab8980eb0f
Author: Doroszlai, Attila <[email protected]>
AuthorDate: Fri Jun 2 09:23:36 2023 +0200
HDDS-8731. Standardize JVM pause monitor (#4814)
---
.../org/apache/hadoop/hdds/ratis/RatisHelper.java | 12 +-
.../common/statemachine/DatanodeStateMachine.java | 36 +---
.../hdds/scm/server/StorageContainerManager.java | 53 +++---
.../org/apache/hadoop/ozone/om/OzoneManager.java | 196 +++++++--------------
.../org/apache/hadoop/ozone/recon/ReconServer.java | 10 ++
.../java/org/apache/hadoop/ozone/s3/Gateway.java | 6 +
6 files changed, 117 insertions(+), 196 deletions(-)
diff --git
a/hadoop-hdds/common/src/main/java/org/apache/hadoop/hdds/ratis/RatisHelper.java
b/hadoop-hdds/common/src/main/java/org/apache/hadoop/hdds/ratis/RatisHelper.java
index 1ad339b7dd..d26e48ea92 100644
---
a/hadoop-hdds/common/src/main/java/org/apache/hadoop/hdds/ratis/RatisHelper.java
+++
b/hadoop-hdds/common/src/main/java/org/apache/hadoop/hdds/ratis/RatisHelper.java
@@ -63,6 +63,7 @@ import org.apache.ratis.rpc.RpcType;
import org.apache.ratis.rpc.SupportedRpcType;
import org.apache.ratis.thirdparty.com.google.protobuf.ByteString;
import org.apache.ratis.thirdparty.io.netty.buffer.ByteBuf;
+import org.apache.ratis.util.JvmPauseMonitor;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
@@ -90,6 +91,10 @@ public final class RatisHelper {
private RatisHelper() {
}
+ public static JvmPauseMonitor newJvmPauseMonitor(String name) {
+ return new JvmPauseMonitor(name, extraSleep -> { });
+ }
+
private static String toRaftPeerIdString(DatanodeDetails id) {
return id.getUuidString();
}
@@ -320,8 +325,6 @@ public final class RatisHelper {
* Set all client properties matching with regex
* {@link RatisHelper#HDDS_DATANODE_RATIS_PREFIX_KEY} in
* ozone configuration object and configure it to RaftProperties.
- * @param ozoneConf
- * @param raftProperties
*/
public static void createRaftClientProperties(ConfigurationSource ozoneConf,
RaftProperties raftProperties) {
@@ -360,8 +363,6 @@ public final class RatisHelper {
* Set all server properties matching with prefix
* {@link RatisHelper#HDDS_DATANODE_RATIS_PREFIX_KEY} in
* ozone configuration object and configure it to RaftProperties.
- * @param ozoneConf
- * @param raftProperties
*/
public static void createRaftServerProperties(ConfigurationSource ozoneConf,
RaftProperties raftProperties) {
@@ -447,7 +448,7 @@ public final class RatisHelper {
}
} catch (IOException e) {
LOG.error("Can not get FirstNode from the pipeline: {} with " +
- "exception: {}", pipeline.toString(), e.getLocalizedMessage());
+ "exception: {}", pipeline, e.getLocalizedMessage());
return null;
}
raftPeers.add(raftPeerId);
@@ -498,7 +499,6 @@ public final class RatisHelper {
*
* @param raftGroup the Raft group
* @param targetPeerId the target expected leader
- * @throws IOException
*/
public static void transferRatisLeadership(ConfigurationSource conf,
RaftGroup raftGroup, RaftPeerId targetPeerId, GrpcTlsConfig tlsConfig)
diff --git
a/hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/statemachine/DatanodeStateMachine.java
b/hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/statemachine/DatanodeStateMachine.java
index 1cfaa44da7..14c6cb9925 100644
---
a/hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/statemachine/DatanodeStateMachine.java
+++
b/hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/statemachine/DatanodeStateMachine.java
@@ -41,7 +41,6 @@ import
org.apache.hadoop.hdds.protocol.proto.StorageContainerDatanodeProtocolPro
import
org.apache.hadoop.hdds.security.x509.certificate.client.CertificateClient;
import org.apache.hadoop.hdds.upgrade.HDDSLayoutVersionManager;
import org.apache.hadoop.hdds.utils.IOUtils;
-import org.apache.hadoop.hdds.utils.LegacyHadoopConfigurationSource;
import org.apache.hadoop.ozone.HddsDatanodeStopService;
import org.apache.hadoop.ozone.container.common.DatanodeLayoutStorage;
import org.apache.hadoop.ozone.container.common.report.ReportManager;
@@ -75,7 +74,6 @@ import
org.apache.hadoop.ozone.container.upgrade.VersionedDatanodeFeatures;
import org.apache.hadoop.ozone.protocol.commands.SCMCommand;
import org.apache.hadoop.ozone.upgrade.UpgradeFinalizer;
import org.apache.hadoop.ozone.upgrade.UpgradeFinalizer.StatusAndMessages;
-import org.apache.hadoop.util.JvmPauseMonitor;
import org.apache.hadoop.util.Time;
import com.google.common.annotations.VisibleForTesting;
@@ -108,13 +106,11 @@ public class DatanodeStateMachine implements Closeable {
private Thread cmdProcessThread = null;
private final ReplicationSupervisor supervisor;
- private JvmPauseMonitor jvmPauseMonitor;
- private CertificateClient dnCertClient;
private final HddsDatanodeStopService hddsDatanodeStopService;
- private HDDSLayoutVersionManager layoutVersionManager;
- private DatanodeLayoutStorage layoutStorage;
- private DataNodeUpgradeFinalizer upgradeFinalizer;
+ private final HDDSLayoutVersionManager layoutVersionManager;
+ private final DatanodeLayoutStorage layoutStorage;
+ private final DataNodeUpgradeFinalizer upgradeFinalizer;
/**
* Used to synchronize to the OzoneContainer object created in the
@@ -126,6 +122,7 @@ public class DatanodeStateMachine implements Closeable {
private final ReplicationSupervisorMetrics replicationSupervisorMetrics;
private final ECReconstructionMetrics ecReconstructionMetrics;
// This is an instance variable as mockito needs to access it in a test
+ @SuppressWarnings("FieldCanBeLocal")
private final ReconstructECContainersCommandHandler
reconstructECContainersCommandHandler;
@@ -176,7 +173,6 @@ public class DatanodeStateMachine implements Closeable {
} finally {
constructionLock.writeLock().unlock();
}
- dnCertClient = certClient;
nextHB = new AtomicLong(Time.monotonicNow());
ContainerImporter importer = new ContainerImporter(conf,
@@ -186,10 +182,10 @@ public class DatanodeStateMachine implements Closeable {
ContainerReplicator pullReplicator = new DownloadAndImportReplicator(
conf, container.getContainerSet(),
importer,
- new SimpleContainerDownloader(conf, dnCertClient));
+ new SimpleContainerDownloader(conf, certClient));
ContainerReplicator pushReplicator = new PushReplicator(conf,
new OnDemandContainerReplicationSource(container.getController()),
- new GrpcContainerUploader(conf, dnCertClient)
+ new GrpcContainerUploader(conf, certClient)
);
pullReplicatorWithMetrics = new MeasuredReplicator(pullReplicator, "pull");
@@ -315,19 +311,13 @@ public class DatanodeStateMachine implements Closeable {
* Runs the state machine at a fixed frequency.
*/
private void startStateMachineThread() throws IOException {
- long now = 0;
+ long now;
reportManager.init();
initCommandHandlerThread(conf);
upgradeFinalizer.runPrefinalizeStateActions(layoutStorage, this);
- // Start jvm monitor
- jvmPauseMonitor = new JvmPauseMonitor();
- jvmPauseMonitor
- .init(LegacyHadoopConfigurationSource.asHadoopConfiguration(conf));
- jvmPauseMonitor.start();
-
while (context.getState() != DatanodeStates.SHUTDOWN) {
try {
LOG.debug("Executing cycle Number : {}", context.getExecutionCount());
@@ -429,10 +419,6 @@ public class DatanodeStateMachine implements Closeable {
container.stop();
}
- if (jvmPauseMonitor != null) {
- jvmPauseMonitor.stop();
- }
-
if (commandDispatcher != null) {
commandDispatcher.stop();
}
@@ -565,8 +551,6 @@ public class DatanodeStateMachine implements Closeable {
/**
* Waits for DatanodeStateMachine to exit.
- *
- * @throws InterruptedException
*/
public void join() throws InterruptedException {
if (stateMachineThread != null) {
@@ -643,12 +627,10 @@ public class DatanodeStateMachine implements Closeable {
/**
* Create a command handler thread.
- *
- * @param config
*/
private void initCommandHandlerThread(ConfigurationSource config) {
- /**
+ /*
* Task that periodically checks if we have any outstanding commands.
* It is assumed that commands can be processed slowly and in order.
* This assumption might change in future. Right now due to this assumption
@@ -657,7 +639,7 @@ public class DatanodeStateMachine implements Closeable {
Runnable processCommandQueue = () -> {
long now;
while (getContext().getState() != DatanodeStates.SHUTDOWN) {
- SCMCommand command = getContext().getNextCommand();
+ SCMCommand<?> command = getContext().getNextCommand();
if (command != null) {
commandDispatcher.handle(command);
commandsHandled++;
diff --git
a/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/StorageContainerManager.java
b/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/StorageContainerManager.java
index 4c8f7bf16e..8bf5b0f1d3 100644
---
a/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/StorageContainerManager.java
+++
b/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/StorageContainerManager.java
@@ -158,9 +158,9 @@ import org.apache.hadoop.security.SecurityUtil;
import org.apache.hadoop.security.UserGroupInformation;
import org.apache.hadoop.security.UserGroupInformation.AuthenticationMethod;
import
org.apache.hadoop.security.authentication.client.AuthenticationException;
-import org.apache.hadoop.util.JvmPauseMonitor;
import org.apache.ratis.protocol.RaftPeerId;
import org.apache.ratis.util.ExitUtils;
+import org.apache.ratis.util.JvmPauseMonitor;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import
org.apache.hadoop.hdds.scm.server.SCMDatanodeHeartbeatDispatcher.ContainerReport;
@@ -188,6 +188,7 @@ import java.util.concurrent.ThreadPoolExecutor;
import java.util.concurrent.TimeUnit;
import static org.apache.hadoop.hdds.HddsUtils.preserveThreadName;
+import static org.apache.hadoop.hdds.ratis.RatisHelper.newJvmPauseMonitor;
import static
org.apache.hadoop.hdds.scm.ScmConfigKeys.OZONE_SCM_EVENT_REPORT_EXEC_WAIT_THRESHOLD_DEFAULT;
import static
org.apache.hadoop.hdds.scm.ScmConfigKeys.OZONE_SCM_EVENT_REPORT_QUEUE_WAIT_THRESHOLD_DEFAULT;
import static
org.apache.hadoop.hdds.security.x509.certificate.authority.CertificateStore.CertType.VALID_CERTS;
@@ -272,7 +273,7 @@ public final class StorageContainerManager extends
ServiceRuntimeInfoImpl
private CertificateClient scmCertificateClient;
private ContainerTokenSecretManager containerTokenMgr;
- private JvmPauseMonitor jvmPauseMonitor;
+ private final JvmPauseMonitor jvmPauseMonitor;
private final OzoneConfiguration configuration;
private SCMContainerMetrics scmContainerMetrics;
private SCMContainerPlacementMetrics placementMetrics;
@@ -283,7 +284,7 @@ public final class StorageContainerManager extends
ServiceRuntimeInfoImpl
private final Map<String, RatisDropwizardExports> ratisMetricsMap =
new ConcurrentHashMap<>();
private List<RatisDropwizardExports.MetricReporter> ratisReporterList = null;
- private String primaryScmNodeId;
+ private final String primaryScmNodeId;
/**
* Network topology Map.
@@ -294,7 +295,7 @@ public final class StorageContainerManager extends
ServiceRuntimeInfoImpl
private final SCMHANodeDetails scmHANodeDetails;
- private ContainerBalancer containerBalancer;
+ private final ContainerBalancer containerBalancer;
// MoveManager is used by ContainerBalancer to schedule container moves
private final MoveManager moveManager;
private StatefulServiceStateManager statefulServiceStateManager;
@@ -342,22 +343,20 @@ public final class StorageContainerManager extends
ServiceRuntimeInfoImpl
Objects.requireNonNull(configurator, "configurator cannot not be null");
Objects.requireNonNull(conf, "configuration cannot not be null");
- /**
- * It is assumed the scm --init command creates the SCM Storage Config.
- */
+ // It is assumed the scm --init command creates the SCM Storage Config.
scmStorageConfig = new SCMStorageConfig(conf);
scmHANodeDetails = SCMHANodeDetails.loadSCMHAConfig(conf,
scmStorageConfig);
configuration = conf;
initMetrics();
+ boolean ratisEnabled = SCMHAUtils.isSCMHAEnabled(conf);
if (scmStorageConfig.getState() != StorageState.INITIALIZED) {
- String errMsg = "Please make sure you have run \'ozone scm --init\' " +
+ String errMsg = "Please make sure you have run 'ozone scm --init' " +
"command to generate all the required metadata to " +
scmStorageConfig.getStorageDir();
- if (SCMHAUtils.isSCMHAEnabled(conf) && !scmStorageConfig
- .isSCMHAEnabled()) {
- errMsg += " or make sure you have run \'ozone scm --bootstrap\' cmd to
"
+ if (ratisEnabled && !scmStorageConfig.isSCMHAEnabled()) {
+ errMsg += " or make sure you have run 'ozone scm --bootstrap' cmd to "
+ "add the SCM to existing SCM HA group";
}
LOG.error(errMsg + ".");
@@ -368,7 +367,9 @@ public final class StorageContainerManager extends
ServiceRuntimeInfoImpl
primaryScmNodeId = scmStorageConfig.getPrimaryScmNodeId();
initializeCertificateClient();
- /**
+ jvmPauseMonitor = !ratisEnabled ? newJvmPauseMonitor(getScmId()) : null;
+
+ /*
* Important : This initialization sequence is assumed by some of our
tests.
* The testSecureOzoneCluster assumes that security checks have to be
* passed before any artifacts like SCM DB is created. So please don't
@@ -613,7 +614,7 @@ public final class StorageContainerManager extends
ServiceRuntimeInfoImpl
}
// This needs to be done before initializing Ratis.
ratisReporterList = RatisDropwizardExports
- .registerRatisMetricReporters(ratisMetricsMap, () -> isStopped.get());
+ .registerRatisMetricReporters(ratisMetricsMap, isStopped::get);
if (configurator.getSCMHAManager() != null) {
scmHAManager = configurator.getSCMHAManager();
} else {
@@ -802,7 +803,6 @@ public final class StorageContainerManager extends
ServiceRuntimeInfoImpl
* @param conf - Config
* @param configurator - configurator
* @throws IOException - on Failure
- * @throws AuthenticationException - on Failure
*/
private void initializeCAnSecurityProtocol(OzoneConfiguration conf,
SCMConfigurator configurator) throws IOException {
@@ -879,9 +879,8 @@ public final class StorageContainerManager extends
ServiceRuntimeInfoImpl
}
}
- /** Persist primary SCM root ca cert and sub-ca certs to DB.
- *
- * @throws IOException
+ /**
+ * Persist primary SCM root ca cert and sub-ca certs to DB.
*/
private void persistPrimarySCMCerts() throws IOException {
BigInteger certSerial =
@@ -994,8 +993,6 @@ public final class StorageContainerManager extends
ServiceRuntimeInfoImpl
/**
* Login as the configured user for SCM.
- *
- * @param conf
*/
private static void loginAsSCMUserIfSecurityEnabled(
SCMHANodeDetails scmhaNodeDetails, ConfigurationSource conf)
@@ -1051,7 +1048,7 @@ public final class StorageContainerManager extends
ServiceRuntimeInfoImpl
public static String buildRpcServerStartMessage(String description,
InetSocketAddress addr) {
return addr != null
- ? String.format("%s is listening at %s", description, addr.toString())
+ ? String.format("%s is listening at %s", description, addr)
: String.format("%s not started", description);
}
@@ -1193,10 +1190,6 @@ public final class StorageContainerManager extends
ServiceRuntimeInfoImpl
/**
* Initialize security If Ozone security is enabled and
* ScmStorageConfig does not have certificate serial id.
- * @param conf
- * @param scmhaNodeDetails
- * @param scmStorageConfig
- * @throws IOException
*/
private static void initializeSecurityIfNeeded(
OzoneConfiguration conf, SCMHANodeDetails scmhaNodeDetails,
@@ -1500,10 +1493,9 @@ public final class StorageContainerManager extends
ServiceRuntimeInfoImpl
scmBlockManager.start();
leaseManager.start();
- // Start jvm monitor
- jvmPauseMonitor = new JvmPauseMonitor();
- jvmPauseMonitor.init(configuration);
- jvmPauseMonitor.start();
+ if (jvmPauseMonitor != null) {
+ jvmPauseMonitor.start();
+ }
try {
httpServer = new StorageContainerManagerHttpServer(configuration, this);
@@ -1519,9 +1511,8 @@ public final class StorageContainerManager extends
ServiceRuntimeInfoImpl
scmHAMetricsUpdate(null);
}
- /** Persist SCM certs to DB on bootstrap scm nodes.
- *
- * @throws IOException
+ /**
+ * Persist SCM certs to DB on bootstrap scm nodes.
*/
private void persistSCMCertificates() throws IOException {
// Fetch all CA's and persist during startup on bootstrap nodes. This
diff --git
a/hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/OzoneManager.java
b/hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/OzoneManager.java
index 263a5b2474..aa31bacad6 100644
---
a/hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/OzoneManager.java
+++
b/hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/OzoneManager.java
@@ -82,6 +82,7 @@ import org.apache.hadoop.hdds.ratis.RatisHelper;
import org.apache.hadoop.hdds.scm.ScmInfo;
import org.apache.hadoop.hdds.scm.client.HddsClientUtils;
import org.apache.hadoop.hdds.server.OzoneAdmins;
+import org.apache.hadoop.hdds.utils.IOUtils;
import org.apache.hadoop.hdds.utils.db.Table;
import org.apache.hadoop.hdds.utils.db.Table.KeyValue;
import org.apache.hadoop.hdds.utils.db.TableIterator;
@@ -209,9 +210,7 @@ import org.apache.hadoop.security.SecurityUtil;
import org.apache.hadoop.security.UserGroupInformation;
import org.apache.hadoop.security.UserGroupInformation.AuthenticationMethod;
import
org.apache.hadoop.security.authentication.client.AuthenticationException;
-import org.apache.hadoop.security.token.SecretManager.InvalidToken;
import org.apache.hadoop.security.token.Token;
-import org.apache.hadoop.util.JvmPauseMonitor;
import org.apache.hadoop.util.KMSUtil;
import org.apache.hadoop.util.Time;
@@ -231,12 +230,11 @@ import static
org.apache.hadoop.hdds.HddsConfigKeys.HDDS_BLOCK_TOKEN_ENABLED;
import static
org.apache.hadoop.hdds.HddsConfigKeys.HDDS_BLOCK_TOKEN_ENABLED_DEFAULT;
import static org.apache.hadoop.hdds.HddsUtils.getScmAddressForClients;
import static org.apache.hadoop.hdds.HddsUtils.preserveThreadName;
+import static org.apache.hadoop.hdds.ratis.RatisHelper.newJvmPauseMonitor;
import static org.apache.hadoop.hdds.server.ServerUtils.updateRPCListenAddress;
import static org.apache.hadoop.hdds.utils.HAUtils.getScmInfo;
import static org.apache.hadoop.ozone.OmUtils.MAX_TRXN_ID;
import static org.apache.hadoop.ozone.OzoneAcl.AclScope.ACCESS;
-import static
org.apache.hadoop.ozone.OzoneConfigKeys.DFS_CONTAINER_RATIS_ENABLED_DEFAULT;
-import static
org.apache.hadoop.ozone.OzoneConfigKeys.DFS_CONTAINER_RATIS_ENABLED_KEY;
import static org.apache.hadoop.ozone.OzoneConfigKeys.OZONE_ACL_ENABLED;
import static
org.apache.hadoop.ozone.OzoneConfigKeys.OZONE_ACL_ENABLED_DEFAULT;
import static org.apache.hadoop.ozone.OzoneConfigKeys.OZONE_ADMINISTRATORS;
@@ -309,6 +307,7 @@ import org.apache.ratis.server.RaftServer;
import org.apache.ratis.server.protocol.TermIndex;
import org.apache.ratis.util.ExitUtils;
import org.apache.ratis.util.FileUtils;
+import org.apache.ratis.util.JvmPauseMonitor;
import org.apache.ratis.util.LifeCycle;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
@@ -343,8 +342,8 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
private OzoneConfiguration configuration;
private RPC.Server omRpcServer;
private GrpcOzoneManagerServer omS3gGrpcServer;
- private InetSocketAddress omRpcAddress;
- private String omId;
+ private final InetSocketAddress omRpcAddress;
+ private final String omId;
private OMMetadataManager metadataManager;
private OMMultiTenantManager multiTenantManager;
@@ -352,7 +351,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
private BucketManager bucketManager;
private KeyManager keyManager;
private PrefixManagerImpl prefixManager;
- private UpgradeFinalizer<OzoneManager> upgradeFinalizer;
+ private final UpgradeFinalizer<OzoneManager> upgradeFinalizer;
/**
* OM super user / admin list.
@@ -378,11 +377,10 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
private static final ObjectReader READER =
new ObjectMapper().readerFor(OmMetricsInfo.class);
private static final int SHUTDOWN_HOOK_PRIORITY = 30;
- private final Runnable shutdownHook;
private final File omMetaDir;
private boolean isAclEnabled;
private final boolean isSpnegoEnabled;
- private JvmPauseMonitor jvmPauseMonitor;
+ private final JvmPauseMonitor jvmPauseMonitor;
private final SecurityConfig secConfig;
private S3SecretManager s3SecretManager;
private final boolean isOmGrpcServerEnabled;
@@ -391,28 +389,26 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
private String omComponent;
private OzoneManagerProtocolServerSideTranslatorPB omServerProtocol;
- private boolean isRatisEnabled;
+ private final boolean isRatisEnabled;
private OzoneManagerRatisServer omRatisServer;
private OmRatisSnapshotProvider omRatisSnapshotProvider;
private OMNodeDetails omNodeDetails;
- private Map<String, OMNodeDetails> peerNodesMap;
+ private final Map<String, OMNodeDetails> peerNodesMap;
private File omRatisSnapshotDir;
private final RatisSnapshotInfo omRatisSnapshotInfo;
private final Map<String, RatisDropwizardExports> ratisMetricsMap =
new ConcurrentHashMap<>();
private List<RatisDropwizardExports.MetricReporter> ratisReporterList = null;
- private KeyProviderCryptoExtension kmsProvider = null;
- private static String keyProviderUriKeyName =
- CommonConfigurationKeysPublic.HADOOP_SECURITY_KEY_PROVIDER_PATH;
- private OMLayoutVersionManager versionManager;
+ private KeyProviderCryptoExtension kmsProvider;
+ private final OMLayoutVersionManager versionManager;
private final ReplicationConfigValidator replicationConfigValidator;
private boolean allowListAllVolumes;
// Adding parameters needed for VolumeRequests here, so that during request
// execution, we can get from ozoneManager.
- private long maxUserVolumeCount;
+ private final long maxUserVolumeCount;
private int minMultipartUploadPartSize = OzoneConsts.OM_MULTIPART_MIN_SIZE;
@@ -420,12 +416,11 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
private final long scmBlockSize;
private final int preallocateBlocksMax;
private final boolean grpcBlockTokenEnabled;
- private final boolean useRatisForReplication;
private final BucketLayout defaultBucketLayout;
- private ReplicationConfig defaultReplicationConfig;
+ private final ReplicationConfig defaultReplicationConfig;
- private boolean isS3MultiTenancyEnabled;
- private boolean isStrictS3;
+ private final boolean isS3MultiTenancyEnabled;
+ private final boolean isStrictS3;
private ExitManager exitManager;
private OzoneManagerPrepareState prepareState;
@@ -439,9 +434,9 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
private static UserGroupInformation testUgi;
private final OzoneLockProvider ozoneLockProvider;
- private OMPerformanceMetrics perfMetrics;
+ private final OMPerformanceMetrics perfMetrics;
- private boolean fsSnapshotEnabled;
+ private final boolean fsSnapshotEnabled;
/**
* OM Startup mode.
@@ -541,8 +536,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
OZONE_KEY_PREALLOCATION_BLOCKS_MAX_DEFAULT);
this.grpcBlockTokenEnabled = conf.getBoolean(HDDS_BLOCK_TOKEN_ENABLED,
HDDS_BLOCK_TOKEN_ENABLED_DEFAULT);
- this.useRatisForReplication = conf.getBoolean(
- DFS_CONTAINER_RATIS_ENABLED_KEY, DFS_CONTAINER_RATIS_ENABLED_DEFAULT);
this.isStrictS3 = conf.getBoolean(
OZONE_OM_NAMESPACE_STRICT_S3,
OZONE_OM_NAMESPACE_STRICT_S3_DEFAULT);
@@ -554,6 +547,9 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
OMConfigKeys.OZONE_OM_RATIS_ENABLE_KEY,
OMConfigKeys.OZONE_OM_RATIS_ENABLE_DEFAULT);
+ // Ratis server comes with JvmPauseMonitor, no need to start another
+ jvmPauseMonitor = !isRatisEnabled ? newJvmPauseMonitor(omId) : null;
+
fsSnapshotEnabled = configuration.getBoolean(
OMConfigKeys.OZONE_FILESYSTEM_SNAPSHOT_ENABLED_KEY,
OMConfigKeys.OZONE_FILESYSTEM_SNAPSHOT_ENABLED_DEFAULT);
@@ -682,10 +678,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
if (isOmGrpcServerEnabled) {
omS3gGrpcServer = getOmS3gGrpcServer(configuration);
}
- shutdownHook = () -> {
- saveOmMetrics();
- };
- ShutdownHookManager.get().addShutdownHook(shutdownHook,
+ ShutdownHookManager.get().addShutdownHook(this::saveOmMetrics,
SHUTDOWN_HOOK_PRIORITY);
if (isBootstrapping || isForcedBootstrapping) {
@@ -748,7 +741,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
private void logVersionMismatch(OzoneConfiguration conf, ScmInfo scmInfo) {
List<SCMNodeInfo> scmNodeInfoList = SCMNodeInfo.buildNodeInfo(conf);
- StringBuilder scmBlockAddressBuilder = new StringBuilder("");
+ StringBuilder scmBlockAddressBuilder = new StringBuilder();
for (SCMNodeInfo scmNodeInfo : scmNodeInfoList) {
scmBlockAddressBuilder.append(scmNodeInfo.getBlockClientAddress())
.append(",");
@@ -851,14 +844,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
}
}
- /**
- * Return configuration value of
- * {@link OzoneConfigKeys#DFS_CONTAINER_RATIS_ENABLED_KEY}.
- */
- public boolean shouldUseRatis() {
- return useRatisForReplication;
- }
-
/**
* Return scmClient.
*/
@@ -934,13 +919,12 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
private KeyProviderCryptoExtension createKeyProviderExt(
OzoneConfiguration conf) throws IOException {
KeyProvider keyProvider = KMSUtil.createKeyProvider(conf,
- keyProviderUriKeyName);
+ CommonConfigurationKeysPublic.HADOOP_SECURITY_KEY_PROVIDER_PATH);
if (keyProvider == null) {
return null;
}
- KeyProviderCryptoExtension cryptoProvider = KeyProviderCryptoExtension
+ return KeyProviderCryptoExtension
.createKeyProviderCryptoExtension(keyProvider);
- return cryptoProvider;
}
@Override
@@ -960,7 +944,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
private void saveOmMetrics() {
try {
- boolean success;
File parent = getTempMetricsStorageFile().getParentFile();
if (!parent.exists()) {
Files.createDirectories(parent.toPath());
@@ -971,14 +954,11 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
OmMetricsInfo metricsInfo = new OmMetricsInfo();
metricsInfo.setNumKeys(metrics.getNumKeys());
WRITER.writeValue(writer, metricsInfo);
- success = true;
}
- if (success) {
- Files.move(getTempMetricsStorageFile().toPath(),
- getMetricsStorageFile().toPath(), StandardCopyOption
- .ATOMIC_MOVE, StandardCopyOption.REPLACE_EXISTING);
- }
+ Files.move(getTempMetricsStorageFile().toPath(),
+ getMetricsStorageFile().toPath(), StandardCopyOption
+ .ATOMIC_MOVE, StandardCopyOption.REPLACE_EXISTING);
} catch (IOException ex) {
LOG.error("Unable to write the om Metrics file", ex);
}
@@ -1127,9 +1107,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
/**
* Login OM service user if security and Kerberos are enabled.
- *
- * @param conf
- * @throws IOException, AuthenticationException
*/
private static void loginOMUser(OzoneConfiguration conf)
throws IOException, AuthenticationException {
@@ -1160,10 +1137,9 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
* Create a scm block client, used by putKey() and getKey().
*
* @return {@link ScmBlockLocationProtocol}
- * @throws IOException
*/
private static ScmBlockLocationProtocol getScmBlockClient(
- OzoneConfiguration conf) throws IOException {
+ OzoneConfiguration conf) {
return HAUtils.getScmBlockClient(conf);
}
@@ -1287,8 +1263,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
* @return gRPC server
* @throws IOException if there is an I/O error while creating RPC server
*/
- private GrpcOzoneManagerServer startGrpcServer(OzoneConfiguration conf)
- throws IOException {
+ private GrpcOzoneManagerServer startGrpcServer(OzoneConfiguration conf) {
return new GrpcOzoneManagerServer(conf,
this.omServerProtocol,
this.delegationTokenMgr,
@@ -1329,7 +1304,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
loginOMUserIfSecurityEnabled(conf);
OMStorage omStorage = new OMStorage(conf);
StorageState state = omStorage.getState();
- String scmId = null;
+ String scmId;
try {
ScmInfo scmInfo = getScmInfo(conf);
scmId = scmInfo.getScmId();
@@ -1386,6 +1361,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
LOG.info("Re-initialize certificate client.");
omStore.unsetOmCertSerialId();
omStore.persistCurrentState();
+ IOUtils.close(LOG, certClient);
certClient = new OMCertificateClient(
new SecurityConfig(conf), omStore, scmId, null, null);
response = certClient.init();
@@ -1484,9 +1460,9 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
*/
private static String buildRpcServerStartMessage(String description,
InetSocketAddress addr) {
- return addr != null ? String.format("%s is listening at %s",
- description, addr.toString()) :
- String.format("%s not started", description);
+ return addr != null
+ ? String.format("%s is listening at %s", description, addr)
+ : String.format("%s not started", description);
}
@VisibleForTesting
@@ -1898,14 +1874,12 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
// NodeIds present in new node list and not in current peer list are the
// bootstapped OMs and should be added to the peer list
- List<String> bootstrappedOMs = new ArrayList<>();
- bootstrappedOMs.addAll(newPeers);
+ List<String> bootstrappedOMs = new ArrayList<>(newPeers);
bootstrappedOMs.removeAll(currentPeers);
// NodeIds present in current peer list but not in new node list are the
// decommissioned OMs and should be removed from the peer list
- List<String> decommissionedOMs = new ArrayList<>();
- decommissionedOMs.addAll(currentPeers);
+ List<String> decommissionedOMs = new ArrayList<>(currentPeers);
decommissionedOMs.removeAll(newPeers);
// Add bootstrapped OMs to peer list
@@ -2070,8 +2044,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
/**
* Starts a Trash Emptier thread that does an fs.trashRoots and performs
* checkpointing & deletion.
- * @param conf
- * @throws IOException
*/
private void startTrashEmptier(Configuration conf) throws IOException {
if (emptier == null) {
@@ -2093,12 +2065,8 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
OzoneManager i = this;
FileSystem fs = SecurityUtil.doAsLoginUser(
- new PrivilegedExceptionAction<FileSystem>() {
- @Override
- public FileSystem run() throws IOException {
- return new TrashOzoneFileSystem(i);
- }
- });
+ (PrivilegedExceptionAction<FileSystem>)
+ () -> new TrashOzoneFileSystem(i));
this.emptier = new Thread(new OzoneTrash(fs, conf, this).
getEmptier(), "Trash Emptier");
this.emptier.setDaemon(true);
@@ -2111,23 +2079,18 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
* for serving s3g OmRequests. If an earlier instance is already running
* then returns the same.
*/
- private GrpcOzoneManagerServer getOmS3gGrpcServer(OzoneConfiguration conf)
- throws IOException {
+ private GrpcOzoneManagerServer getOmS3gGrpcServer(OzoneConfiguration conf) {
if (isOmGrpcServerRunning) {
return omS3gGrpcServer;
}
return startGrpcServer(configuration);
}
- /**
- * Creates an instance of ratis server.
- */
/**
* Creates an instance of ratis server.
* @param shouldBootstrap If OM is started in Bootstrap mode, then Ratis
* server will be initialized without adding self to
* Ratis group
- * @throws IOException
*/
private void initializeRatisServer(boolean shouldBootstrap)
throws IOException {
@@ -2135,7 +2098,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
if (omRatisServer == null) {
// This needs to be done before initializing Ratis.
ratisReporterList = RatisDropwizardExports.
- registerRatisMetricReporters(ratisMetricsMap, () -> isStopped());
+ registerRatisMetricReporters(ratisMetricsMap, this::isStopped);
omRatisServer = OzoneManagerRatisServer.newOMRatisServer(
configuration, this, omNodeDetails, peerNodesMap,
secConfig, certClient, shouldBootstrap);
@@ -2302,20 +2265,16 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
*/
private boolean isAllowedDelegationTokenOp() throws IOException {
AuthenticationMethod authMethod = getConnectionAuthenticationMethod();
- if (UserGroupInformation.isSecurityEnabled()
- && (authMethod != AuthenticationMethod.KERBEROS)
- && (authMethod != AuthenticationMethod.KERBEROS_SSL)
- && (authMethod != AuthenticationMethod.CERTIFICATE)) {
- return false;
- }
- return true;
+ return !UserGroupInformation.isSecurityEnabled()
+ || (authMethod == AuthenticationMethod.KERBEROS)
+ || (authMethod == AuthenticationMethod.KERBEROS_SSL)
+ || (authMethod == AuthenticationMethod.CERTIFICATE);
}
/**
* Returns authentication method used to establish the connection.
*
* @return AuthenticationMethod used to establish connection
- * @throws IOException
*/
private AuthenticationMethod getConnectionAuthenticationMethod()
throws IOException {
@@ -2339,12 +2298,10 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
*
* @param renewer Renewer information
* @return delegationToken DelegationToken signed by OzoneManager
- * @throws IOException on error
*/
@Override
public Token<OzoneTokenIdentifier> getDelegationToken(Text renewer)
throws OMException {
- Token<OzoneTokenIdentifier> token;
try {
if (!isAllowedDelegationTokenOp()) {
throw new OMException("Delegation Token can be issued only with "
@@ -2379,8 +2336,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
*
* @param token token to renew
* @return new expiryTime of the token
- * @throws InvalidToken if {@code token} is invalid
- * @throws IOException on other errors
*/
@Override
public long renewDelegationToken(Token<OzoneTokenIdentifier> token)
@@ -2403,7 +2358,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
OzoneTokenIdentifier id = null;
try {
id = OzoneTokenIdentifier.readProtoBuf(token.getIdentifier());
- } catch (IOException exe) {
+ } catch (IOException ignored) {
}
LOG.error("Delegation token renewal failed for dt id: {}, cause: {}",
id, ex.getMessage());
@@ -2417,7 +2372,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
* Cancels a delegation token.
*
* @param token token to cancel
- * @throws IOException on error
*/
@Override
public void cancelDelegationToken(Token<OzoneTokenIdentifier> token)
@@ -2441,11 +2395,8 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
if (ownerName == null) {
return false;
}
- if (callerUgi.getUserName().equals(ownerName) ||
- callerUgi.getShortUserName().equals(ownerName)) {
- return true;
- }
- return false;
+ return callerUgi.getUserName().equals(ownerName) ||
+ callerUgi.getShortUserName().equals(ownerName);
}
/**
@@ -2481,7 +2432,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
Boolean lockAcquired = metadataManager.getLock().acquireReadLock(
VOLUME_LOCK, volume);
String dbVolumeKey = metadataManager.getVolumeKey(volume);
- OmVolumeArgs volumeArgs = null;
+ OmVolumeArgs volumeArgs;
try {
volumeArgs = metadataManager.getVolumeTable().get(dbVolumeKey);
} catch (IOException ioe) {
@@ -2521,7 +2472,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
private String getBucketOwner(String volume, String bucket)
throws OMException {
- OmBucketInfo bucketInfo = null;
+ OmBucketInfo bucketInfo;
boolean lockAcquired = metadataManager.getLock().acquireReadLock(
BUCKET_LOCK, volume, bucket);
@@ -2612,7 +2563,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
*
* @param volume - Volume name.
* @return VolumeArgs or exception is thrown.
- * @throws IOException
*/
@Override
public OmVolumeArgs getVolumeInfo(String volume) throws IOException {
@@ -2649,7 +2599,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
* prevkey
* @param maxKeys - Max number of keys to return.
* @return List of Volumes.
- * @throws IOException
*/
@Override
public List<OmVolumeArgs> listVolumeByUser(String userName, String prefix,
@@ -2717,7 +2666,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
* prevkey
* @param maxKeys - Max number of keys to return.
* @return List of Volumes.
- * @throws IOException
*/
@Override
public List<OmVolumeArgs> listAllVolumes(String prefix, String prevKey, int
@@ -2798,7 +2746,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
* @param volume - Volume name.
* @param bucket - Bucket name.
* @return OmBucketInfo or exception is thrown.
- * @throws IOException
*/
@Override
public OmBucketInfo getBucketInfo(String volume, String bucket)
@@ -2813,9 +2760,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
bucket, null);
}
metrics.incNumBucketInfos();
- final OmBucketInfo bucketInfo =
- bucketManager.getBucketInfo(volume, bucket);
- return bucketInfo;
+ return bucketManager.getBucketInfo(volume, bucket);
} catch (Exception ex) {
metrics.incNumBucketInfoFails();
auditSuccess = false;
@@ -2835,7 +2780,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
*
* @param args - attributes of the key.
* @return OmKeyInfo - the info about the requested key.
- * @throws IOException
*/
@Override
public OmKeyInfo lookupKey(OmKeyArgs args) throws IOException {
@@ -2961,7 +2905,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
@Override
public String getRatisRoles() {
- List<ServiceInfo> serviceList = null;
+ List<ServiceInfo> serviceList;
int port = omNodeDetails.getRatisPort();
RaftPeer leaderId;
if (isRatisEnabled) {
@@ -2974,7 +2918,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
serviceList = getServiceList();
} catch (IOException e) {
LOG.error("IO-Exception Occurred", e);
- return "Exception: " + e.toString();
+ return "Exception: " + e;
}
return OmUtils.format(serviceList, port, leaderId.getId().toString());
} else {
@@ -3787,7 +3731,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
* @param lastAppliedIndex the last applied index in the current OM DB.
* @param checkpointPath path to the new DB checkpoint
* @return location of backup of the original DB
- * @throws Exception
*/
File replaceOMDBWithCheckpoint(long lastAppliedIndex, File oldDB,
Path checkpointPath) throws IOException {
@@ -3988,19 +3931,16 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
* process the requests.
*
* If ratis is not enabled, then it always returns true.
- * @return
*/
public boolean isLeaderReady() {
- return isRatisEnabled ?
- omRatisServer.checkLeaderStatus() == LEADER_AND_READY : true;
+ return !isRatisEnabled
+ || omRatisServer.checkLeaderStatus() == LEADER_AND_READY;
}
/**
- * Check the leader status.
- *
- * @return null leader is ready
- * OMLeaderNotReadyException leader is not ready
- * OMNotLeaderException not leader
+ * Checks the leader status. Does nothing if this OM is leader and is ready.
+ * @throws OMLeaderNotReadyException if leader, but not ready
+ * @throws OMNotLeaderException if not leader
*/
public void checkLeaderStatus() throws OMNotLeaderException,
OMLeaderNotReadyException {
@@ -4027,8 +3967,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
/**
* Return if Ratis is enabled or not.
- *
- * @return
*/
public boolean isRatisEnabled() {
return isRatisEnabled;
@@ -4124,10 +4062,9 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
}
private void startJVMPauseMonitor() {
- // Start jvm monitor
- jvmPauseMonitor = new JvmPauseMonitor();
- jvmPauseMonitor.init(configuration);
- jvmPauseMonitor.start();
+ if (jvmPauseMonitor != null) {
+ jvmPauseMonitor.start();
+ }
}
public ResolvedBucket resolveBucketLink(KeyArgs args,
@@ -4185,9 +4122,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
* @param volumeAndBucket the bucket to be resolved (if it is a link)
* @param visited collects link buckets visited during the resolution to
* avoid infinite loops
- * @param {@link UserGroupInformation}
- * @param remoteAddress
- * @param hostName
* @return bucket location possibly updated with its actual volume and bucket
* after following bucket links
* @throws IOException (most likely OMException) if ACL check fails, bucket
is
@@ -4265,7 +4199,6 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
/**
* Create volume which is required for S3Gateway operations.
- * @throws IOException
*/
private void addS3GVolumeToDB() throws IOException {
String s3VolumeName =
HddsClientUtils.getDefaultS3VolumeName(configuration);
@@ -4349,7 +4282,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
List<String> userGroups = Arrays.asList(UserGroupInformation
.createRemoteUser(userName).getGroupNames());
- userGroups.stream().forEach((group) -> listOfAcls.add(
+ userGroups.forEach((group) -> listOfAcls.add(
new OzoneAcl(ACLIdentityType.GROUP, group, ACLType.ALL, ACCESS)));
// Add ACLs
@@ -4476,14 +4409,13 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
@Override
public EchoRPCResponse echoRPCReq(byte[] payloadReq,
- int payloadSizeResp)
- throws IOException {
+ int payloadSizeResp) {
return null;
}
@Override
public boolean recoverLease(String volumeName, String bucketName,
- String keyName) throws IOException {
+ String keyName) {
return false;
}
@@ -4600,7 +4532,7 @@ public final class OzoneManager extends
ServiceRuntimeInfoImpl
}
@Override // ReconfigureProtocol
- public String getServerName() throws IOException {
+ public String getServerName() {
return "OM";
}
diff --git
a/hadoop-ozone/recon/src/main/java/org/apache/hadoop/ozone/recon/ReconServer.java
b/hadoop-ozone/recon/src/main/java/org/apache/hadoop/ozone/recon/ReconServer.java
index d8e77cb3cc..bdd81a3a72 100644
---
a/hadoop-ozone/recon/src/main/java/org/apache/hadoop/ozone/recon/ReconServer.java
+++
b/hadoop-ozone/recon/src/main/java/org/apache/hadoop/ozone/recon/ReconServer.java
@@ -46,6 +46,7 @@ import org.apache.hadoop.ozone.util.ShutdownHookManager;
import org.apache.hadoop.security.SecurityUtil;
import org.apache.hadoop.security.UserGroupInformation;
import
org.apache.hadoop.security.authentication.client.AuthenticationException;
+import org.apache.ratis.util.JvmPauseMonitor;
import org.hadoop.ozone.recon.codegen.ReconSchemaGenerationModule;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
@@ -53,6 +54,7 @@ import org.slf4j.LoggerFactory;
import java.io.IOException;
import java.net.InetSocketAddress;
+import static org.apache.hadoop.hdds.ratis.RatisHelper.newJvmPauseMonitor;
import static
org.apache.hadoop.hdds.recon.ReconConfig.ConfigStrings.OZONE_RECON_KERBEROS_KEYTAB_FILE_KEY;
import static
org.apache.hadoop.hdds.recon.ReconConfig.ConfigStrings.OZONE_RECON_KERBEROS_PRINCIPAL_KEY;
import static org.apache.hadoop.ozone.common.Storage.StorageState.INITIALIZED;
@@ -67,6 +69,7 @@ public class ReconServer extends GenericCli {
private static final Logger LOG = LoggerFactory.getLogger(ReconServer.class);
private Injector injector;
+ private JvmPauseMonitor jvmPauseMonitor;
private ReconHttpServer httpServer;
private ReconContainerMetadataManager reconContainerMetadataManager;
private OzoneManagerServiceProvider ozoneManagerServiceProvider;
@@ -121,6 +124,7 @@ public class ReconServer extends GenericCli {
} catch (Exception e) {
LOG.error("Error during initializing Recon certificate", e);
}
+ jvmPauseMonitor = newJvmPauseMonitor("Recon");
this.reconDBProvider = injector.getInstance(ReconDBProvider.class);
this.reconContainerMetadataManager =
injector.getInstance(ReconContainerMetadataManager.class);
@@ -246,6 +250,9 @@ public class ReconServer extends GenericCli {
if (reconStorageContainerManager != null) {
reconStorageContainerManager.start();
}
+ if (jvmPauseMonitor != null) {
+ jvmPauseMonitor.start();
+ }
}
}
@@ -290,6 +297,9 @@ public class ReconServer extends GenericCli {
LOG.error("Failed to close certificate client.", ioe);
}
}
+ if (jvmPauseMonitor != null) {
+ jvmPauseMonitor.stop();
+ }
}
public void join() {
diff --git
a/hadoop-ozone/s3gateway/src/main/java/org/apache/hadoop/ozone/s3/Gateway.java
b/hadoop-ozone/s3gateway/src/main/java/org/apache/hadoop/ozone/s3/Gateway.java
index 9865d4d7fd..02bf8ff1f4 100644
---
a/hadoop-ozone/s3gateway/src/main/java/org/apache/hadoop/ozone/s3/Gateway.java
+++
b/hadoop-ozone/s3gateway/src/main/java/org/apache/hadoop/ozone/s3/Gateway.java
@@ -39,11 +39,13 @@ import org.apache.hadoop.ozone.util.ShutdownHookManager;
import org.apache.hadoop.security.SecurityUtil;
import org.apache.hadoop.security.UserGroupInformation;
import
org.apache.hadoop.security.authentication.client.AuthenticationException;
+import org.apache.ratis.util.JvmPauseMonitor;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import picocli.CommandLine.Command;
import static org.apache.hadoop.hdds.StringUtils.startupShutdownMessage;
+import static org.apache.hadoop.hdds.ratis.RatisHelper.newJvmPauseMonitor;
import static
org.apache.hadoop.ozone.conf.OzoneServiceConfig.DEFAULT_SHUTDOWN_HOOK_PRIORITY;
import static
org.apache.hadoop.ozone.s3.S3GatewayConfigKeys.OZONE_S3G_KERBEROS_KEYTAB_FILE_KEY;
import static
org.apache.hadoop.ozone.s3.S3GatewayConfigKeys.OZONE_S3G_KERBEROS_PRINCIPAL_KEY;
@@ -63,6 +65,8 @@ public class Gateway extends GenericCli {
private S3GatewayMetrics metrics;
private OzoneConfiguration ozoneConfiguration;
+ private final JvmPauseMonitor jvmPauseMonitor = newJvmPauseMonitor("S3G");
+
public static void main(String[] args) throws Exception {
OzoneNetUtils.disableJvmNetworkAddressCacheIfRequired(
new OzoneConfiguration());
@@ -120,12 +124,14 @@ public class Gateway extends GenericCli {
LOG.info("Starting Ozone S3 gateway");
HddsServerUtil.initializeMetrics(ozoneConfiguration, "S3Gateway");
+ jvmPauseMonitor.start();
httpServer.start();
}
public void stop() throws Exception {
LOG.info("Stopping Ozone S3 gateway");
httpServer.stop();
+ jvmPauseMonitor.stop();
S3GatewayMetrics.unRegister();
}
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
