HDDS-1214. Enable tracing for the datanode read/write path. Contributed by Elek, Marton.

Closes #550.
This commit is contained in:
Márton Elek 2019-03-04 15:25:32 +01:00
parent c7307867f0
commit d17e31e062
No known key found for this signature in database
GPG Key ID: D51EA8F00EE79B28
6 changed files with 88 additions and 54 deletions

View File

@ -22,6 +22,7 @@
import io.jaegertracing.Configuration;
import io.jaegertracing.internal.JaegerTracer;
import io.opentracing.Scope;
import io.opentracing.Span;
import io.opentracing.SpanContext;
import io.opentracing.Tracer;
import io.opentracing.util.GlobalTracer;
@ -64,6 +65,19 @@ public static String exportCurrentSpan() {
return builder.toString();
}
/**
* Export the specific span as a string.
*
* @return encoded tracing context.
*/
public static String exportSpan(Span span) {
StringBuilder builder = new StringBuilder();
if (span != null) {
GlobalTracer.get().inject(span.context(), StringCodec.FORMAT, builder);
}
return builder.toString();
}
/**
* Create a new scope and use the imported span as the parent.
*

View File

@ -33,6 +33,7 @@
.InvalidContainerStateException;
import org.apache.hadoop.hdds.scm.container.common.helpers
.StorageContainerException;
import org.apache.hadoop.hdds.tracing.TracingUtil;
import org.apache.hadoop.ozone.audit.AuditAction;
import org.apache.hadoop.ozone.audit.AuditEventStatus;
import org.apache.hadoop.ozone.audit.AuditLogger;
@ -61,6 +62,8 @@
ContainerDataProto.State;
import org.apache.hadoop.hdds.protocol.datanode.proto.ContainerProtos.Result;
import org.apache.hadoop.ozone.container.common.interfaces.ContainerDispatcher;
import io.opentracing.Scope;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
@ -137,10 +140,19 @@ public void buildMissingContainerSet(Set<Long> createdContainerSet) {
containerSet.buildMissingContainerSet(createdContainerSet);
}
@SuppressWarnings("methodlength")
@Override
public ContainerCommandResponseProto dispatch(
ContainerCommandRequestProto msg, DispatcherContext dispatcherContext) {
String spanName = "HddsDispatcher." + msg.getCmdType().name();
try (Scope scope = TracingUtil
.importAndCreateScope(spanName, msg.getTraceID())) {
return dispatchRequest(msg, dispatcherContext);
}
}
@SuppressWarnings("methodlength")
private ContainerCommandResponseProto dispatchRequest(
ContainerCommandRequestProto msg, DispatcherContext dispatcherContext) {
Preconditions.checkNotNull(msg);
LOG.trace("Command {}, trace ID: {} ", msg.getCmdType().toString(),
msg.getTraceID());

View File

@ -24,6 +24,7 @@
.StorageContainerDatanodeProtocolProtos.SCMCommandProto;
import org.apache.hadoop.hdds.protocol.proto
.StorageContainerDatanodeProtocolProtos.CloseContainerCommandProto;
import org.apache.hadoop.hdds.tracing.TracingUtil;
import org.apache.hadoop.ozone.container.common.interfaces.Container;
import org.apache.hadoop.ozone.container.common.statemachine
.SCMConnectionManager;
@ -133,6 +134,7 @@ private ContainerCommandRequestProto getContainerCommandRequestProto(
final ContainerCommandRequestProto.Builder command =
ContainerCommandRequestProto.newBuilder();
command.setCmdType(ContainerProtos.Type.CloseContainer);
command.setTraceID(TracingUtil.exportCurrentSpan());
command.setContainerID(containerId);
command.setCloseContainer(
ContainerProtos.CloseContainerRequestProto.getDefaultInstance());

View File

@ -31,10 +31,12 @@
import org.apache.hadoop.hdds.scm.container.common.helpers.
StorageContainerException;
import org.apache.hadoop.hdds.tracing.GrpcServerInterceptor;
import org.apache.hadoop.hdds.tracing.TracingUtil;
import org.apache.hadoop.ozone.OzoneConfigKeys;
import org.apache.hadoop.ozone.OzoneConsts;
import org.apache.hadoop.ozone.container.common.interfaces.ContainerDispatcher;
import io.opentracing.Scope;
import org.apache.ratis.thirdparty.io.grpc.BindableService;
import org.apache.ratis.thirdparty.io.grpc.Server;
import org.apache.ratis.thirdparty.io.grpc.ServerBuilder;
@ -168,12 +170,18 @@ public void stop() {
@Override
public void submitRequest(ContainerCommandRequestProto request,
HddsProtos.PipelineID pipelineID) throws IOException {
super.submitRequest(request, pipelineID);
ContainerProtos.ContainerCommandResponseProto response =
storageContainer.dispatch(request, null);
if (response.getResult() != ContainerProtos.Result.SUCCESS) {
throw new StorageContainerException(response.getMessage(),
response.getResult());
try (Scope scope = TracingUtil
.importAndCreateScope(
"XceiverServerGrpc." + request.getCmdType().name(),
request.getTraceID())) {
super.submitRequest(request, pipelineID);
ContainerProtos.ContainerCommandResponseProto response =
storageContainer.dispatch(request, null);
if (response.getResult() != ContainerProtos.Result.SUCCESS) {
throw new StorageContainerException(response.getMessage(),
response.getResult());
}
}
}

View File

@ -25,7 +25,6 @@
import org.apache.hadoop.hdds.HddsUtils;
import org.apache.hadoop.hdds.protocol.datanode.proto.ContainerProtos;
import io.opentracing.Scope;
import org.apache.hadoop.hdds.scm.container.common.helpers.StorageContainerException;
import org.apache.hadoop.ozone.container.common.helpers.ContainerUtils;
import org.apache.ratis.proto.RaftProtos.RaftPeerRole;
@ -51,7 +50,6 @@
.ReadChunkRequestProto;
import org.apache.hadoop.hdds.protocol.datanode.proto.ContainerProtos
.ReadChunkResponseProto;
import org.apache.hadoop.hdds.tracing.TracingUtil;
import org.apache.hadoop.ozone.container.common.interfaces.ContainerDispatcher;
import org.apache.hadoop.hdds.security.token.TokenVerifier;
import org.apache.hadoop.security.UserGroupInformation;
@ -271,51 +269,50 @@ public TransactionContext startTransaction(RaftClientRequest request)
final ContainerCommandRequestProto proto =
getContainerCommandRequestProto(request.getMessage().getContent());
Preconditions.checkArgument(request.getRaftGroupId().equals(gid));
try (Scope scope = TracingUtil
.importAndCreateScope(proto.getCmdType().name(), proto.getTraceID())) {
try {
dispatcher.validateContainerCommand(proto);
} catch (IOException ioe) {
TransactionContext ctxt = TransactionContext.newBuilder()
.setClientRequest(request)
.setStateMachine(this)
.setServerRole(RaftPeerRole.LEADER)
.build();
ctxt.setException(ioe);
return ctxt;
}
if (proto.getCmdType() == Type.WriteChunk) {
final WriteChunkRequestProto write = proto.getWriteChunk();
// create the log entry proto
final WriteChunkRequestProto commitWriteChunkProto =
WriteChunkRequestProto.newBuilder()
.setBlockID(write.getBlockID())
.setChunkData(write.getChunkData())
// skipping the data field as it is
// already set in statemachine data proto
.build();
ContainerCommandRequestProto commitContainerCommandProto =
ContainerCommandRequestProto
.newBuilder(proto)
.setWriteChunk(commitWriteChunkProto)
.build();
return TransactionContext.newBuilder()
.setClientRequest(request)
.setStateMachine(this)
.setServerRole(RaftPeerRole.LEADER)
.setStateMachineData(write.getData())
.setLogData(commitContainerCommandProto.toByteString())
.build();
} else {
return TransactionContext.newBuilder()
.setClientRequest(request)
.setStateMachine(this)
.setServerRole(RaftPeerRole.LEADER)
.setLogData(request.getMessage().getContent())
.build();
}
try {
dispatcher.validateContainerCommand(proto);
} catch (IOException ioe) {
TransactionContext ctxt = TransactionContext.newBuilder()
.setClientRequest(request)
.setStateMachine(this)
.setServerRole(RaftPeerRole.LEADER)
.build();
ctxt.setException(ioe);
return ctxt;
}
if (proto.getCmdType() == Type.WriteChunk) {
final WriteChunkRequestProto write = proto.getWriteChunk();
// create the log entry proto
final WriteChunkRequestProto commitWriteChunkProto =
WriteChunkRequestProto.newBuilder()
.setBlockID(write.getBlockID())
.setChunkData(write.getChunkData())
// skipping the data field as it is
// already set in statemachine data proto
.build();
ContainerCommandRequestProto commitContainerCommandProto =
ContainerCommandRequestProto
.newBuilder(proto)
.setWriteChunk(commitWriteChunkProto)
.setTraceID(proto.getTraceID())
.build();
return TransactionContext.newBuilder()
.setClientRequest(request)
.setStateMachine(this)
.setServerRole(RaftPeerRole.LEADER)
.setStateMachineData(write.getData())
.setLogData(commitContainerCommandProto.toByteString())
.build();
} else {
return TransactionContext.newBuilder()
.setClientRequest(request)
.setStateMachine(this)
.setServerRole(RaftPeerRole.LEADER)
.setLogData(request.getMessage().getContent())
.build();
}
}
private ByteString getStateMachineData(StateMachineLogEntryProto entryProto) {

View File

@ -486,7 +486,8 @@ public void submitRequest(ContainerCommandRequestProto request,
super.submitRequest(request, pipelineID);
RaftClientReply reply;
try (Scope scope = TracingUtil
.importAndCreateScope(request.getCmdType().name(),
.importAndCreateScope(
"XceiverServerRatis." + request.getCmdType().name(),
request.getTraceID())) {
RaftClientRequest raftClientRequest =