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

HDDS-806. Update Ratis to latest snapshot version in ozone. Contributed by Tsz Wo Nicholas Sze and Mukul Kumar Singh.

This commit is contained in:
Shashikant Banerjee 2018-11-09 00:05:45 +05:30
parent 8d99648c20
commit 31614bcc7c
7 changed files with 67 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
hadoop-hdds/common/src/main/java/org/apache/hadoop/hdds/scm/ScmConfigKeys.java
View File

@ -85,6 +85,14 @@ public final class ScmConfigKeys {
public static final TimeDuration
DFS_CONTAINER_RATIS_STATEMACHINEDATA_SYNC_TIMEOUT_DEFAULT =
TimeDuration.valueOf(10, TimeUnit.SECONDS);
public static final String
DFS_CONTAINER_RATIS_STATEMACHINEDATA_SYNC_RETRIES =
"dfs.container.ratis.statemachinedata.sync.retries";
public static final int
DFS_CONTAINER_RATIS_STATEMACHINEDATA_SYNC_RETRIES_DEFAULT = -1;
public static final String DFS_CONTAINER_RATIS_LOG_QUEUE_SIZE =
"dfs.container.ratis.log.queue.size";
public static final int DFS_CONTAINER_RATIS_LOG_QUEUE_SIZE_DEFAULT = 128;
public static final String DFS_RATIS_CLIENT_REQUEST_TIMEOUT_DURATION_KEY =
"dfs.ratis.client.request.timeout.duration";
public static final TimeDuration

10
hadoop-hdds/common/src/main/java/org/apache/hadoop/ozone/OzoneConfigKeys.java
View File

@ -261,6 +261,16 @@ public final class OzoneConfigKeys {
public static final TimeDuration
DFS_RATIS_SERVER_RETRY_CACHE_TIMEOUT_DURATION_DEFAULT =
ScmConfigKeys.DFS_RATIS_SERVER_RETRY_CACHE_TIMEOUT_DURATION_DEFAULT;
public static final String
DFS_CONTAINER_RATIS_STATEMACHINEDATA_SYNC_RETRIES =
ScmConfigKeys.DFS_CONTAINER_RATIS_STATEMACHINEDATA_SYNC_RETRIES;
public static final int
DFS_CONTAINER_RATIS_STATEMACHINEDATA_SYNC_RETRIES_DEFAULT =
ScmConfigKeys.DFS_CONTAINER_RATIS_STATEMACHINEDATA_SYNC_RETRIES_DEFAULT;
public static final String DFS_CONTAINER_RATIS_LOG_QUEUE_SIZE =
ScmConfigKeys.DFS_CONTAINER_RATIS_LOG_QUEUE_SIZE;
public static final int DFS_CONTAINER_RATIS_LOG_QUEUE_SIZE_DEFAULT =
ScmConfigKeys.DFS_CONTAINER_RATIS_LOG_QUEUE_SIZE_DEFAULT;
public static final String DFS_RATIS_SERVER_REQUEST_TIMEOUT_DURATION_KEY =
ScmConfigKeys.DFS_RATIS_SERVER_REQUEST_TIMEOUT_DURATION_KEY;
public static final TimeDuration

15
hadoop-hdds/common/src/main/resources/ozone-default.xml
View File

@ -59,6 +59,21 @@
<description>Timeout for StateMachine data writes by Ratis.
</description>
</property>
<property>
<name>dfs.container.ratis.statemachinedata.sync.retries</name>
<value>-1</value>
<tag>OZONE, DEBUG, CONTAINER, RATIS</tag>
<description>Number of times the WriteStateMachineData op will be tried
before failing, if this value is -1, then this retries indefinitely.
</description>
</property>
<property>
<name>dfs.container.ratis.log.queue.size</name>
<value>128</value>
<tag>OZONE, DEBUG, CONTAINER, RATIS</tag>
<description>Number of operation pending with Raft's Log Worker.
</description>
</property>
<property>
<name>dfs.container.ratis.datanode.storage.dir</name>
<value/>

48
hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/transport/server/ratis/ContainerStateMachine.java
View File

@ -22,7 +22,7 @@ import com.google.common.base.Preconditions;
import org.apache.hadoop.hdds.HddsUtils;
import org.apache.hadoop.hdds.protocol.datanode.proto.ContainerProtos;
import org.apache.hadoop.ozone.container.common.helpers.BlockData;
import org.apache.ratis.proto.RaftProtos.StateMachineEntryProto;
import org.apache.ratis.proto.RaftProtos.RaftPeerRole;
import org.apache.ratis.protocol.RaftGroup;
import org.apache.ratis.protocol.RaftGroupId;
import org.apache.ratis.server.RaftServer;
@ -55,7 +55,6 @@ import org.apache.ratis.statemachine.StateMachineStorage;
import org.apache.ratis.statemachine.TransactionContext;
import org.apache.ratis.statemachine.impl.BaseStateMachine;
import org.apache.ratis.statemachine.impl.SimpleStateMachineStorage;
import org.apache.ratis.statemachine.impl.TransactionContextImpl;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
@ -206,7 +205,6 @@ public class ContainerStateMachine extends BaseStateMachine {
final ContainerCommandRequestProto proto =
getRequestProto(request.getMessage().getContent());
Preconditions.checkArgument(request.getRaftGroupId().equals(gid));
final StateMachineLogEntryProto log;
if (proto.getCmdType() == Type.WriteChunk) {
final WriteChunkRequestProto write = proto.getWriteChunk();
// create the state machine data proto
@ -236,33 +234,29 @@ public class ContainerStateMachine extends BaseStateMachine {
.setWriteChunk(commitWriteChunkProto)
.build();
log = createSMLogEntryProto(request,
commitContainerCommandProto.toByteString(),
dataContainerCommandProto.toByteString());
return TransactionContext.newBuilder()
.setClientRequest(request)
.setStateMachine(this)
.setServerRole(RaftPeerRole.LEADER)
.setStateMachineData(dataContainerCommandProto.toByteString())
.setLogData(commitContainerCommandProto.toByteString())
.build();
} else if (proto.getCmdType() == Type.CreateContainer) {
log = createSMLogEntryProto(request,
request.getMessage().getContent(), request.getMessage().getContent());
return TransactionContext.newBuilder()
.setClientRequest(request)
.setStateMachine(this)
.setServerRole(RaftPeerRole.LEADER)
.setStateMachineData(request.getMessage().getContent())
.setLogData(request.getMessage().getContent())
.build();
} else {
log = createSMLogEntryProto(request, request.getMessage().getContent(),
null);
return TransactionContext.newBuilder()
.setClientRequest(request)
.setStateMachine(this)
.setServerRole(RaftPeerRole.LEADER)
.setLogData(request.getMessage().getContent())
.build();
}
return new TransactionContextImpl(this, request, log);
}
private StateMachineLogEntryProto createSMLogEntryProto(RaftClientRequest r,
ByteString logData, ByteString smData) {
StateMachineLogEntryProto.Builder builder =
StateMachineLogEntryProto.newBuilder();
builder.setCallId(r.getCallId())
.setClientId(r.getClientId().toByteString())
.setLogData(logData);
if (smData != null) {
builder.setStateMachineEntry(StateMachineEntryProto.newBuilder()
.setStateMachineData(smData).build());
}
return builder.build();
}
private ByteString getStateMachineData(StateMachineLogEntryProto entryProto) {

11
hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/transport/server/ratis/XceiverServerRatis.java
View File

@ -287,6 +287,17 @@ public final class XceiverServerRatis implements XceiverServerSpi {
setAutoTriggerEnabled(properties, true);
RaftServerConfigKeys.Snapshot.
setAutoTriggerThreshold(properties, snapshotThreshold);
int logQueueSize =
conf.getInt(OzoneConfigKeys.DFS_CONTAINER_RATIS_LOG_QUEUE_SIZE,
OzoneConfigKeys.DFS_CONTAINER_RATIS_LOG_QUEUE_SIZE_DEFAULT);
RaftServerConfigKeys.Log.setQueueSize(properties, logQueueSize);
int numSyncRetries = conf.getInt(
OzoneConfigKeys.DFS_CONTAINER_RATIS_STATEMACHINEDATA_SYNC_RETRIES,
OzoneConfigKeys.
DFS_CONTAINER_RATIS_STATEMACHINEDATA_SYNC_RETRIES_DEFAULT);
RaftServerConfigKeys.Log.StateMachineData.setSyncTimeoutRetry(properties,
numSyncRetries);
return properties;
}

2
hadoop-hdds/pom.xml
View File

@ -45,7 +45,7 @@ http://maven.apache.org/xsd/maven-4.0.0.xsd">
<hdds.version>0.4.0-SNAPSHOT</hdds.version>
<!-- Apache Ratis version -->
<ratis.version>0.3.0-1d2ebee-SNAPSHOT</ratis.version>
<ratis.version>0.3.0-1d07b18-SNAPSHOT</ratis.version>
<bouncycastle.version>1.60</bouncycastle.version>

2
hadoop-ozone/pom.xml
View File

@ -33,7 +33,7 @@ http://maven.apache.org/xsd/maven-4.0.0.xsd">
<hadoop.version>3.2.1-SNAPSHOT</hadoop.version>
<hdds.version>0.4.0-SNAPSHOT</hdds.version>
<ozone.version>0.4.0-SNAPSHOT</ozone.version>
<ratis.version>0.3.0-1d2ebee-SNAPSHOT</ratis.version>
<ratis.version>0.3.0-1d07b18-SNAPSHOT</ratis.version>
<bouncycastle.version>1.60</bouncycastle.version>
<ozone.release>Badlands</ozone.release>
<declared.ozone.version>${ozone.version}</declared.ozone.version>