import akka.actor.ActorRef;
import com.google.common.annotations.VisibleForTesting;
import com.google.common.base.Preconditions;
-import com.google.common.base.Verify;
import edu.umd.cs.findbugs.annotations.SuppressFBWarnings;
-import java.util.ArrayDeque;
-import java.util.Iterator;
-import java.util.Map.Entry;
import java.util.Optional;
-import java.util.Queue;
import java.util.concurrent.TimeUnit;
+import java.util.concurrent.locks.Lock;
+import java.util.concurrent.locks.ReentrantLock;
import java.util.function.Consumer;
import javax.annotation.Nonnull;
import javax.annotation.concurrent.GuardedBy;
import javax.annotation.concurrent.NotThreadSafe;
import org.opendaylight.controller.cluster.access.concepts.Request;
-import org.opendaylight.controller.cluster.access.concepts.RequestEnvelope;
import org.opendaylight.controller.cluster.access.concepts.RequestException;
import org.opendaylight.controller.cluster.access.concepts.Response;
import org.opendaylight.controller.cluster.access.concepts.ResponseEnvelope;
@VisibleForTesting
static final long REQUEST_TIMEOUT_NANOS = TimeUnit.SECONDS.toNanos(30);
- private final Queue<TransmittedConnectionEntry> inflight;
- private final Queue<ConnectionEntry> pending;
-
+ private final Lock lock = new ReentrantLock();
private final ClientActorContext context;
+ @GuardedBy("lock")
+ private final TransmitQueue queue;
private final Long cookie;
- private volatile ReconnectForwarder successor;
private volatile RequestException poisoned;
private long lastProgress;
- private AbstractClientConnection(final ClientActorContext context, final Long cookie,
- final Queue<TransmittedConnectionEntry> inflight, final Queue<ConnectionEntry> pending) {
+ // Do not allow subclassing outside of this package
+ AbstractClientConnection(final ClientActorContext context, final Long cookie,
+ final TransmitQueue queue) {
this.context = Preconditions.checkNotNull(context);
this.cookie = Preconditions.checkNotNull(cookie);
- this.inflight = Preconditions.checkNotNull(inflight);
- this.pending = Preconditions.checkNotNull(pending);
+ this.queue = Preconditions.checkNotNull(queue);
this.lastProgress = readTime();
}
- // Do not allow subclassing outside of this package
- AbstractClientConnection(final ClientActorContext context, final Long cookie) {
- this(context, cookie, new ArrayDeque<>(), new ArrayDeque<>(1));
- }
-
// Do not allow subclassing outside of this package
AbstractClientConnection(final AbstractClientConnection<T> oldConnection) {
- this(oldConnection.context, oldConnection.cookie, oldConnection.inflight, oldConnection.pending);
+ this.context = oldConnection.context;
+ this.cookie = oldConnection.cookie;
+ this.lastProgress = oldConnection.lastProgress;
+ this.queue = new TransmitQueue.Halted();
}
public final ClientActorContext context() {
* @param callback Callback to invoke
*/
public final void sendRequest(final Request<?, ?> request, final Consumer<Response<?, ?>> callback) {
- Preconditions.checkState(poisoned == null, "Connection %s has been poisoned", this);
-
- final ReconnectForwarder beforeQueue = successor;
- final ConnectionEntry entry = new ConnectionEntry(request, callback, readTime());
- if (beforeQueue != null) {
- LOG.trace("Forwarding entry {} from {} to {}", entry, this, beforeQueue);
- beforeQueue.forwardEntry(entry);
- return;
+ final RequestException maybePoison = poisoned;
+ if (maybePoison != null) {
+ throw new IllegalStateException("Connection " + this + " has been poisoned", maybePoison);
}
- enqueueEntry(entry);
+ final ConnectionEntry entry = new ConnectionEntry(request, callback, readTime());
- final ReconnectForwarder afterQueue = successor;
- if (afterQueue != null) {
- synchronized (this) {
- spliceToSuccessor(afterQueue);
- }
+ lock.lock();
+ try {
+ queue.enqueue(entry, entry.getEnqueuedTicks());
+ } finally {
+ lock.unlock();
}
}
- public final synchronized void setForwarder(final ReconnectForwarder forwarder) {
- Verify.verify(successor == null, "Successor {} already set on connection {}", successor, this);
- successor = Preconditions.checkNotNull(forwarder);
- LOG.debug("Connection {} superseded by {}, splicing queue", this, successor);
- spliceToSuccessor(forwarder);
- }
-
public abstract Optional<T> getBackendInfo();
- abstract ClientActorBehavior<T> reconnectConnection(ClientActorBehavior<T> current);
-
- abstract int remoteMaxMessages();
-
- abstract Entry<ActorRef, RequestEnvelope> prepareForTransmit(Request<?, ?> req);
-
- @GuardedBy("this")
- final void spliceToSuccessor(final ReconnectForwarder successor) {
- ConnectionEntry entry = inflight.poll();
- while (entry != null) {
- successor.forwardEntry(entry);
- entry = inflight.poll();
- }
-
- entry = pending.poll();
- while (entry != null) {
- successor.forwardEntry(entry);
- entry = pending.poll();
- }
+ final Iterable<ConnectionEntry> startReplay() {
+ lock.lock();
+ return queue.asIterable();
}
- private long readTime() {
- return context.ticker().read();
+ @GuardedBy("lock")
+ final void finishReplay(final ReconnectForwarder forwarder) {
+ queue.setForwarder(forwarder, readTime());
+ lock.unlock();
}
- private void transmit(final ConnectionEntry entry) {
- final Entry<ActorRef, RequestEnvelope> tuple = prepareForTransmit(entry.getRequest());
- final RequestEnvelope req = tuple.getValue();
+ @GuardedBy("lock")
+ final void setForwarder(final ReconnectForwarder forwarder) {
+ queue.setForwarder(forwarder, readTime());
+ }
- // We need to enqueue the request before we send it to the actor, as we may be executing on a different thread
- // than the client actor thread, in which case the round-trip could be made faster than we can enqueue --
- // in which case the receive routine would not find the entry.
- final TransmittedConnectionEntry txEntry = new TransmittedConnectionEntry(entry, req.getSessionId(),
- req.getTxSequence(), readTime());
- inflight.add(txEntry);
+ @GuardedBy("lock")
+ abstract ClientActorBehavior<T> reconnectConnection(ClientActorBehavior<T> current);
- final ActorRef actor = tuple.getKey();
- LOG.trace("Transmitting request {} as {} to {}", entry.getRequest(), req, actor);
- actor.tell(req, ActorRef.noSender());
+ private long readTime() {
+ return context.ticker().read();
}
- final void enqueueEntry(final ConnectionEntry entry) {
- if (inflight.size() < remoteMaxMessages()) {
- transmit(entry);
- LOG.debug("Enqueued request {} to queue {}", entry.getRequest(), this);
- } else {
- LOG.debug("Queue is at capacity, delayed sending of request {}", entry.getRequest());
- pending.add(entry);
+ final void enqueueEntry(final ConnectionEntry entry, final long now) {
+ lock.lock();
+ try {
+ queue.enqueue(entry, now);
+ } finally {
+ lock.unlock();
}
}
*/
@VisibleForTesting
final ClientActorBehavior<T> runTimer(final ClientActorBehavior<T> current) {
- final long now = readTime();
-
- if (!inflight.isEmpty() || !pending.isEmpty()) {
- final long ticksSinceProgress = now - lastProgress;
- if (ticksSinceProgress >= NO_PROGRESS_TIMEOUT_NANOS) {
- LOG.error("Queue {} has not seen progress in {} seconds, failing all requests", this,
- TimeUnit.NANOSECONDS.toSeconds(ticksSinceProgress));
-
- poison(new NoProgressException(ticksSinceProgress));
- current.removeConnection(this);
- return current;
+ final Optional<FiniteDuration> delay;
+
+ lock.lock();
+ try {
+ final long now = readTime();
+ if (!queue.isEmpty()) {
+ final long ticksSinceProgress = now - lastProgress;
+ if (ticksSinceProgress >= NO_PROGRESS_TIMEOUT_NANOS) {
+ LOG.error("Queue {} has not seen progress in {} seconds, failing all requests", this,
+ TimeUnit.NANOSECONDS.toSeconds(ticksSinceProgress));
+
+ lockedPoison(new NoProgressException(ticksSinceProgress));
+ current.removeConnection(this);
+ return current;
+ }
}
- }
- // Requests are always scheduled in sequence, hence checking for timeout is relatively straightforward.
- // Note we use also inquire about the delay, so we can re-schedule if needed, hence the unusual tri-state
- // return convention.
- final Optional<FiniteDuration> delay = checkTimeout(now);
- if (delay == null) {
- // We have timed out. There is no point in scheduling a timer
- return reconnectConnection(current);
+ // Requests are always scheduled in sequence, hence checking for timeout is relatively straightforward.
+ // Note we use also inquire about the delay, so we can re-schedule if needed, hence the unusual tri-state
+ // return convention.
+ delay = lockedCheckTimeout(now);
+ if (delay == null) {
+ // We have timed out. There is no point in scheduling a timer
+ return reconnectConnection(current);
+ }
+ } finally {
+ lock.unlock();
}
if (delay.isPresent()) {
return current;
}
+ @VisibleForTesting
+ final Optional<FiniteDuration> checkTimeout(final long now) {
+ lock.lock();
+ try {
+ return lockedCheckTimeout(now);
+ } finally {
+ lock.unlock();
+ }
+ }
+
/*
* We are using tri-state return here to indicate one of three conditions:
* - if there is no timeout to schedule, return Optional.empty()
*/
@SuppressFBWarnings(value = "NP_OPTIONAL_RETURN_NULL",
justification = "Returning null Optional is documented in the API contract.")
- private Optional<FiniteDuration> checkTimeout(final ConnectionEntry head, final long now) {
+ @GuardedBy("lock")
+ private Optional<FiniteDuration> lockedCheckTimeout(final long now) {
+ final ConnectionEntry head = queue.peek();
if (head == null) {
return Optional.empty();
}
return Optional.of(FiniteDuration.apply(delay, TimeUnit.NANOSECONDS));
}
- /*
- * We are using tri-state return here to indicate one of three conditions:
- * - if there is no timeout to schedule, return Optional.empty()
- * - if there is a timeout to schedule, return a non-empty optional
- * - if this connections has timed out, return null
- */
- @SuppressFBWarnings(value = "NP_OPTIONAL_RETURN_NULL",
- justification = "Returning null Optional is documented in the API contract.")
- @VisibleForTesting
- final Optional<FiniteDuration> checkTimeout(final long now) {
- final Optional<FiniteDuration> xmit = checkTimeout(inflight.peek(), now);
- if (xmit == null) {
- return null;
- }
- final Optional<FiniteDuration> pend = checkTimeout(pending.peek(), now);
- if (pend == null) {
- return null;
- }
- if (!xmit.isPresent()) {
- return pend;
- }
- if (!pend.isPresent()) {
- return xmit;
+ final void poison(final RequestException cause) {
+ lock.lock();
+ try {
+ lockedPoison(cause);
+ } finally {
+ lock.unlock();
}
-
- return Optional.of(xmit.get().min(pend.get()));
}
- final void poison(final RequestException cause) {
+ @GuardedBy("lock")
+ private void lockedPoison(final RequestException cause) {
poisoned = cause;
-
- poisonQueue(inflight, cause);
- poisonQueue(pending, cause);
+ queue.poison(cause);
}
@VisibleForTesting
}
final void receiveResponse(final ResponseEnvelope<?> envelope) {
- Optional<TransmittedConnectionEntry> maybeEntry = findMatchingEntry(inflight, envelope);
- if (maybeEntry == null) {
- LOG.debug("Request for {} not found in inflight queue, checking pending queue", envelope);
- maybeEntry = findMatchingEntry(pending, envelope);
- }
-
- if (maybeEntry == null || !maybeEntry.isPresent()) {
- LOG.warn("No request matching {} found, ignoring response", envelope);
- return;
- }
-
- final TransmittedConnectionEntry entry = maybeEntry.get();
- LOG.debug("Completing {} with {}", entry, envelope);
- entry.complete(envelope.getMessage());
-
- // We have freed up a slot, try to transmit something
- int toSend = remoteMaxMessages() - inflight.size();
- while (toSend > 0) {
- final ConnectionEntry e = pending.poll();
- if (e == null) {
- break;
- }
+ final long now = readTime();
- LOG.debug("Transmitting entry {}", e);
- transmit(e);
- toSend--;
+ lock.lock();
+ try {
+ queue.complete(envelope, now);
+ } finally {
+ lock.unlock();
}
lastProgress = readTime();
}
-
- private static void poisonQueue(final Queue<? extends ConnectionEntry> queue, final RequestException cause) {
- for (ConnectionEntry e : queue) {
- final Request<?, ?> request = e.getRequest();
- LOG.trace("Poisoning request {}", request, cause);
- e.complete(request.toRequestFailure(cause));
- }
- queue.clear();
- }
-
- /*
- * We are using tri-state return here to indicate one of three conditions:
- * - if a matching entry is found, return an Optional containing it
- * - if a matching entry is not found, but it makes sense to keep looking at other queues, return null
- * - if a conflicting entry is encountered, indicating we should ignore this request, return an empty Optional
- */
- @SuppressFBWarnings(value = "NP_OPTIONAL_RETURN_NULL",
- justification = "Returning null Optional is documented in the API contract.")
- private static Optional<TransmittedConnectionEntry> findMatchingEntry(final Queue<? extends ConnectionEntry> queue,
- final ResponseEnvelope<?> envelope) {
- // Try to find the request in a queue. Responses may legally come back in a different order, hence we need
- // to use an iterator
- final Iterator<? extends ConnectionEntry> it = queue.iterator();
- while (it.hasNext()) {
- final ConnectionEntry e = it.next();
- final Request<?, ?> request = e.getRequest();
- final Response<?, ?> response = envelope.getMessage();
-
- // First check for matching target, or move to next entry
- if (!request.getTarget().equals(response.getTarget())) {
- continue;
- }
-
- // Sanity-check logical sequence, ignore any out-of-order messages
- if (request.getSequence() != response.getSequence()) {
- LOG.debug("Expecting sequence {}, ignoring response {}", request.getSequence(), envelope);
- return Optional.empty();
- }
-
- // Check if the entry has (ever) been transmitted
- if (!(e instanceof TransmittedConnectionEntry)) {
- return Optional.empty();
- }
-
- final TransmittedConnectionEntry te = (TransmittedConnectionEntry) e;
-
- // Now check session match
- if (envelope.getSessionId() != te.getSessionId()) {
- LOG.debug("Expecting session {}, ignoring response {}", te.getSessionId(), envelope);
- return Optional.empty();
- }
- if (envelope.getTxSequence() != te.getTxSequence()) {
- LOG.warn("Expecting txSequence {}, ignoring response {}", te.getTxSequence(), envelope);
- return Optional.empty();
- }
-
- LOG.debug("Completing request {} with {}", request, envelope);
- it.remove();
- return Optional.of(te);
- }
-
- return null;
- }
}
*/
abstract class AbstractReceivingClientConnection<T extends BackendInfo> extends AbstractClientConnection<T> {
private final T backend;
- private long nextTxSequence;
AbstractReceivingClientConnection(final ClientActorContext context, final Long cookie, final T backend) {
- super(context, cookie);
+ super(context, cookie, new TransmitQueue.Transmitting(backend));
this.backend = Preconditions.checkNotNull(backend);
}
AbstractReceivingClientConnection(final AbstractReceivingClientConnection<T> oldConnection) {
super(oldConnection);
this.backend = oldConnection.backend;
- this.nextTxSequence = oldConnection.nextTxSequence;
}
@Override
final T backend() {
return backend;
}
-
- final long nextTxSequence() {
- return nextTxSequence++;
- }
}
import com.google.common.annotations.Beta;
import com.google.common.base.Preconditions;
+import com.google.common.base.Verify;
import java.util.Map;
import java.util.concurrent.ConcurrentHashMap;
import javax.annotation.Nonnull;
import org.opendaylight.controller.cluster.access.concepts.RequestFailure;
import org.opendaylight.controller.cluster.access.concepts.ResponseEnvelope;
import org.opendaylight.controller.cluster.access.concepts.RetiredGenerationException;
+import org.opendaylight.controller.cluster.access.concepts.RuntimeRequestException;
import org.opendaylight.controller.cluster.access.concepts.SuccessEnvelope;
import org.opendaylight.controller.cluster.access.concepts.TransactionIdentifier;
import org.opendaylight.yangtools.concepts.Identifiable;
@Beta
public abstract class ClientActorBehavior<T extends BackendInfo> extends
RecoveredClientActorBehavior<ClientActorContext> implements Identifiable<ClientIdentifier> {
+ /**
+ * Connection reconnect cohort, driven by this class.
+ */
+ @FunctionalInterface
+ protected interface ConnectionConnectCohort {
+ /**
+ * Finish the connection by replaying previous messages onto the new connection.
+ *
+ * @param enqueuedEntries Previously-enqueued entries
+ * @return A {@link ReconnectForwarder} to handle any straggler messages which arrive after this method returns.
+ */
+ @Nonnull ReconnectForwarder finishReconnect(@Nonnull Iterable<ConnectionEntry> enqueuedEntries);
+ }
+
private static final Logger LOG = LoggerFactory.getLogger(ClientActorBehavior.class);
/**
}
/**
- * Callback invoked when a new connection has been established.
+ * Callback invoked when a new connection has been established. Implementations are expected perform preparatory
+ * tasks before the previous connection is frozen.
*
- * @param conn Old connection
- * @param backend New backend
- * @return Newly-connected connection.
+ * @param newConn New connection
+ * @return ConnectionConnectCohort which will be used to complete the process of bringing the connection up.
*/
@GuardedBy("connectionsLock")
- protected abstract @Nonnull ConnectedClientConnection<T> connectionUp(
- final @Nonnull AbstractClientConnection<T> conn, final @Nonnull T backend);
+ @Nonnull protected abstract ConnectionConnectCohort connectionUp(@Nonnull ConnectedClientConnection<T> newConn);
private void backendConnectFinished(final Long shard, final AbstractClientConnection<T> conn,
final T backend, final Throwable failure) {
if (failure != null) {
LOG.error("{}: failed to resolve shard {}", persistenceId(), shard, failure);
+ conn.poison(new RuntimeRequestException("Failed to resolve shard " + shard, failure));
return;
}
LOG.debug("{}: resolved shard {} to {}", persistenceId(), shard, backend);
final long stamp = connectionsLock.writeLock();
try {
- // Bring the connection up
- final ConnectedClientConnection<T> newConn = connectionUp(conn, backend);
+ // Create a new connected connection
+ final ConnectedClientConnection<T> newConn = new ConnectedClientConnection<>(conn.context(),
+ conn.cookie(), backend);
+ LOG.debug("{}: resolving connection {} to {}", persistenceId(), conn, newConn);
+
+ // Start reconnecting without the old connection lock held
+ final ConnectionConnectCohort cohort = Verify.verifyNotNull(connectionUp(newConn));
+
+ // Lock the old connection and get a reference to its entries
+ final Iterable<ConnectionEntry> replayIterable = conn.startReplay();
+
+ // Finish the connection attempt
+ final ReconnectForwarder forwarder = Verify.verifyNotNull(cohort.finishReconnect(replayIterable));
+
+ // Install the forwarder, unlocking the old connection
+ conn.finishReplay(forwarder);
// Make sure new lookups pick up the new connection
connections.replace(shard, conn, newConn);
*/
package org.opendaylight.controller.cluster.access.client;
-import akka.actor.ActorRef;
import com.google.common.annotations.Beta;
-import java.util.AbstractMap.SimpleImmutableEntry;
-import java.util.Map.Entry;
import javax.annotation.concurrent.NotThreadSafe;
-import org.opendaylight.controller.cluster.access.concepts.Request;
-import org.opendaylight.controller.cluster.access.concepts.RequestEnvelope;
@Beta
@NotThreadSafe
public final class ConnectedClientConnection<T extends BackendInfo> extends AbstractReceivingClientConnection<T> {
- public ConnectedClientConnection(final ClientActorContext context, final Long cookie, final T backend) {
+ ConnectedClientConnection(final ClientActorContext context, final Long cookie, final T backend) {
super(context, cookie, backend);
}
current.reconnectConnection(this, next);
return current;
}
-
- @Override
- int remoteMaxMessages() {
- return backend().getMaxMessages();
- }
-
- @Override
- Entry<ActorRef, RequestEnvelope> prepareForTransmit(final Request<?, ?> req) {
- return new SimpleImmutableEntry<>(backend().getActor(), new RequestEnvelope(
- req.toVersion(backend().getVersion()), backend().getSessionId(), nextTxSequence()));
- }
}
*/
package org.opendaylight.controller.cluster.access.client;
-import akka.actor.ActorRef;
import com.google.common.annotations.Beta;
-import java.util.Map.Entry;
import java.util.Optional;
-import org.opendaylight.controller.cluster.access.concepts.Request;
-import org.opendaylight.controller.cluster.access.concepts.RequestEnvelope;
@Beta
public final class ConnectingClientConnection<T extends BackendInfo> extends AbstractClientConnection<T> {
// Initial state, never instantiated externally
ConnectingClientConnection(final ClientActorContext context, final Long cookie) {
- super(context, cookie);
+ super(context, cookie, new TransmitQueue.Halted());
}
@Override
ClientActorBehavior<T> reconnectConnection(final ClientActorBehavior<T> current) {
throw new UnsupportedOperationException("Attempted to reconnect a connecting connection");
}
-
- @Override
- Entry<ActorRef, RequestEnvelope> prepareForTransmit(final Request<?, ?> req) {
- // This is guarded by remoteMaxMessages() == 0
- throw new UnsupportedOperationException("Attempted to transmit on a connecting connection");
- }
-
- @Override
- int remoteMaxMessages() {
- return 0;
- }
}
successor.sendRequest(request, callback);
}
- protected abstract void forwardEntry(ConnectionEntry entry);
+ protected abstract void forwardEntry(ConnectionEntry entry, long now);
final AbstractReceivingClientConnection<?> successor() {
return successor;
*/
package org.opendaylight.controller.cluster.access.client;
-import akka.actor.ActorRef;
-import java.util.Map.Entry;
-import org.opendaylight.controller.cluster.access.concepts.Request;
-import org.opendaylight.controller.cluster.access.concepts.RequestEnvelope;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
LOG.debug("Skipping reconnect of already-reconnecting connection {}", this);
return current;
}
-
- @Override
- Entry<ActorRef, RequestEnvelope> prepareForTransmit(final Request<?, ?> req) {
- // This is guarded by remoteMaxMessages() == 0
- throw new UnsupportedOperationException("Attempted to transmit on a reconnecting connection");
- }
-
- @Override
- int remoteMaxMessages() {
- return 0;
- }
}
}
@Override
- protected void forwardEntry(final ConnectionEntry entry) {
- successor().enqueueEntry(entry);
+ protected void forwardEntry(final ConnectionEntry entry, final long now) {
+ successor().enqueueEntry(entry, now);
}
}
--- /dev/null
+/*
+ * Copyright (c) 2016 Cisco Systems, Inc. and others. All rights reserved.
+ *
+ * This program and the accompanying materials are made available under the
+ * terms of the Eclipse Public License v1.0 which accompanies this distribution,
+ * and is available at http://www.eclipse.org/legal/epl-v10.html
+ */
+package org.opendaylight.controller.cluster.access.client;
+
+import akka.actor.ActorRef;
+import com.google.common.base.Preconditions;
+import com.google.common.base.Verify;
+import com.google.common.collect.Iterables;
+import edu.umd.cs.findbugs.annotations.SuppressFBWarnings;
+import java.util.ArrayDeque;
+import java.util.Iterator;
+import java.util.Optional;
+import java.util.Queue;
+import javax.annotation.concurrent.NotThreadSafe;
+import org.opendaylight.controller.cluster.access.concepts.Request;
+import org.opendaylight.controller.cluster.access.concepts.RequestEnvelope;
+import org.opendaylight.controller.cluster.access.concepts.RequestException;
+import org.opendaylight.controller.cluster.access.concepts.Response;
+import org.opendaylight.controller.cluster.access.concepts.ResponseEnvelope;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+/**
+ * This queue is internally split into two queues for performance reasons, both memory efficiency and copy
+ * operations.
+ *
+ * <p>
+ * Entries are always appended to the end, but then they are transmitted to the remote end and do not necessarily
+ * complete in the order in which they were sent -- hence the head of the queue does not increase linearly,
+ * but can involve spurious removals of non-head entries.
+ *
+ * <p>
+ * For memory efficiency we want to pre-allocate both queues -- which points to ArrayDeque, but that is very
+ * inefficient when entries are removed from the middle. In the typical case we expect the number of in-flight
+ * entries to be an order of magnitude lower than the number of enqueued entries, hence the split.
+ *
+ * <p>
+ * Note that in transient case of reconnect, when the backend gives us a lower number of maximum in-flight entries
+ * than the previous incarnation, we may end up still moving the pending queue -- but that is a very exceptional
+ * scenario, hence we consciously ignore it to keep the design relatively simple.
+ *
+ * <p>
+ * This class is not thread-safe, as it is expected to be guarded by {@link AbstractClientConnection}.
+ *
+ * @author Robert Varga
+ */
+@NotThreadSafe
+abstract class TransmitQueue {
+ static final class Halted extends TransmitQueue {
+ @Override
+ int canTransmitCount(final int inflightSize) {
+ return 0;
+ }
+
+ @Override
+ TransmittedConnectionEntry transmit(final ConnectionEntry entry, final long now) {
+ throw new UnsupportedOperationException("Attempted to transmit on a halted queue");
+ }
+ }
+
+ static final class Transmitting extends TransmitQueue {
+ private final BackendInfo backend;
+ private long nextTxSequence;
+
+ Transmitting(final BackendInfo backend) {
+ this.backend = Preconditions.checkNotNull(backend);
+ }
+
+ @Override
+ int canTransmitCount(final int inflightSize) {
+ return backend.getMaxMessages() - inflightSize;
+ }
+
+ @Override
+ TransmittedConnectionEntry transmit(final ConnectionEntry entry, final long now) {
+ final RequestEnvelope env = new RequestEnvelope(entry.getRequest().toVersion(backend.getVersion()),
+ backend.getSessionId(), nextTxSequence++);
+
+ final TransmittedConnectionEntry ret = new TransmittedConnectionEntry(entry, env.getSessionId(),
+ env.getTxSequence(), now);
+ backend.getActor().tell(env, ActorRef.noSender());
+ return ret;
+ }
+ }
+
+ private static final Logger LOG = LoggerFactory.getLogger(TransmitQueue.class);
+
+ private final ArrayDeque<TransmittedConnectionEntry> inflight = new ArrayDeque<>();
+ private final ArrayDeque<ConnectionEntry> pending = new ArrayDeque<>();
+
+ private ReconnectForwarder successor;
+
+ final Iterable<ConnectionEntry> asIterable() {
+ return Iterables.concat(inflight, pending);
+ }
+
+ private void recordCompletion(final long now, final long enqueuedTicks, final long transmitTicks,
+ final long execNanos) {
+ // TODO: record
+ }
+
+ final void complete(final ResponseEnvelope<?> envelope, final long now) {
+ Optional<TransmittedConnectionEntry> maybeEntry = findMatchingEntry(inflight, envelope);
+ if (maybeEntry == null) {
+ LOG.debug("Request for {} not found in inflight queue, checking pending queue", envelope);
+ maybeEntry = findMatchingEntry(pending, envelope);
+ }
+
+ if (maybeEntry == null || !maybeEntry.isPresent()) {
+ LOG.warn("No request matching {} found, ignoring response", envelope);
+ return;
+ }
+
+ final TransmittedConnectionEntry entry = maybeEntry.get();
+ LOG.debug("Completing {} with {}", entry, envelope);
+ entry.complete(envelope.getMessage());
+
+ recordCompletion(now, entry.getEnqueuedTicks(), entry.getTxTicks(), envelope.getExecutionTimeNanos());
+
+ // We have freed up a slot, try to transmit something
+ int toSend = canTransmitCount(inflight.size());
+ while (toSend > 0) {
+ final ConnectionEntry e = pending.poll();
+ if (e == null) {
+ break;
+ }
+
+ LOG.debug("Transmitting entry {}", e);
+ transmit(e, now);
+ toSend--;
+ }
+ }
+
+ final void enqueue(final ConnectionEntry entry, final long now) {
+ if (successor != null) {
+ successor.forwardEntry(entry, now);
+ return;
+ }
+
+ if (canTransmitCount(inflight.size()) <= 0) {
+ LOG.trace("Queue is at capacity, delayed sending of request {}", entry.getRequest());
+ pending.add(entry);
+ return;
+ }
+
+ // We are not thread-safe and are supposed to be externally-guarded, hence send-before-record should be fine.
+ // This needs to be revisited if the external guards are lowered.
+ inflight.offer(transmit(entry, now));
+ LOG.debug("Sent request {} on queue {}", entry.getRequest(), this);
+ }
+
+ abstract int canTransmitCount(int inflightSize);
+
+ abstract TransmittedConnectionEntry transmit(ConnectionEntry entry, long now);
+
+ final boolean isEmpty() {
+ return inflight.isEmpty() && pending.isEmpty();
+ }
+
+ final ConnectionEntry peek() {
+ final ConnectionEntry ret = inflight.peek();
+ if (ret != null) {
+ return ret;
+ }
+
+ return pending.peek();
+ }
+
+ final void poison(final RequestException cause) {
+ poisonQueue(inflight, cause);
+ poisonQueue(pending, cause);
+ }
+
+ final void setForwarder(final ReconnectForwarder forwarder, final long now) {
+ Verify.verify(successor == null, "Successor {} already set on connection {}", successor, this);
+ successor = Preconditions.checkNotNull(forwarder);
+ LOG.debug("Connection {} superseded by {}, splicing queue", this, successor);
+
+ ConnectionEntry entry = inflight.poll();
+ while (entry != null) {
+ successor.forwardEntry(entry, now);
+ entry = inflight.poll();
+ }
+
+ entry = pending.poll();
+ while (entry != null) {
+ successor.forwardEntry(entry, now);
+ entry = pending.poll();
+ }
+ }
+
+ /*
+ * We are using tri-state return here to indicate one of three conditions:
+ * - if a matching entry is found, return an Optional containing it
+ * - if a matching entry is not found, but it makes sense to keep looking at other queues, return null
+ * - if a conflicting entry is encountered, indicating we should ignore this request, return an empty Optional
+ */
+ @SuppressFBWarnings(value = "NP_OPTIONAL_RETURN_NULL",
+ justification = "Returning null Optional is documented in the API contract.")
+ private static Optional<TransmittedConnectionEntry> findMatchingEntry(final Queue<? extends ConnectionEntry> queue,
+ final ResponseEnvelope<?> envelope) {
+ // Try to find the request in a queue. Responses may legally come back in a different order, hence we need
+ // to use an iterator
+ final Iterator<? extends ConnectionEntry> it = queue.iterator();
+ while (it.hasNext()) {
+ final ConnectionEntry e = it.next();
+ final Request<?, ?> request = e.getRequest();
+ final Response<?, ?> response = envelope.getMessage();
+
+ // First check for matching target, or move to next entry
+ if (!request.getTarget().equals(response.getTarget())) {
+ continue;
+ }
+
+ // Sanity-check logical sequence, ignore any out-of-order messages
+ if (request.getSequence() != response.getSequence()) {
+ LOG.debug("Expecting sequence {}, ignoring response {}", request.getSequence(), envelope);
+ return Optional.empty();
+ }
+
+ // Check if the entry has (ever) been transmitted
+ if (!(e instanceof TransmittedConnectionEntry)) {
+ return Optional.empty();
+ }
+
+ final TransmittedConnectionEntry te = (TransmittedConnectionEntry) e;
+
+ // Now check session match
+ if (envelope.getSessionId() != te.getSessionId()) {
+ LOG.debug("Expecting session {}, ignoring response {}", te.getSessionId(), envelope);
+ return Optional.empty();
+ }
+ if (envelope.getTxSequence() != te.getTxSequence()) {
+ LOG.warn("Expecting txSequence {}, ignoring response {}", te.getTxSequence(), envelope);
+ return Optional.empty();
+ }
+
+ LOG.debug("Completing request {} with {}", request, envelope);
+ it.remove();
+ return Optional.of(te);
+ }
+
+ return null;
+ }
+
+ private static void poisonQueue(final Queue<? extends ConnectionEntry> queue, final RequestException cause) {
+ for (ConnectionEntry e : queue) {
+ final Request<?, ?> request = e.getRequest();
+ LOG.trace("Poisoning request {}", request, cause);
+ e.complete(request.toRequestFailure(cause));
+ }
+ queue.clear();
+ }
+
+}
import javax.annotation.concurrent.GuardedBy;
import org.opendaylight.controller.cluster.access.client.AbstractClientConnection;
import org.opendaylight.controller.cluster.access.client.ConnectedClientConnection;
+import org.opendaylight.controller.cluster.access.client.ConnectionEntry;
import org.opendaylight.controller.cluster.access.client.InversibleLockException;
import org.opendaylight.controller.cluster.access.commands.CreateLocalHistoryRequest;
import org.opendaylight.controller.cluster.access.concepts.LocalHistoryIdentifier;
}
@Override
- void replaySuccessfulRequests() {
- proxy.replaySuccessfulRequests();
+ void replaySuccessfulRequests(final Iterable<ConnectionEntry> previousEntries) {
+ proxy.replaySuccessfulRequests(previousEntries);
}
@Override
import java.util.Map;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.atomic.AtomicLong;
-import javax.annotation.concurrent.GuardedBy;
-import org.opendaylight.controller.cluster.access.client.AbstractClientConnection;
import org.opendaylight.controller.cluster.access.client.BackendInfoResolver;
import org.opendaylight.controller.cluster.access.client.ClientActorBehavior;
import org.opendaylight.controller.cluster.access.client.ClientActorContext;
* involved, as the messages need to be replayed to the individual proxies.
*/
@Override
- @GuardedBy("connectionsLock")
- protected final ConnectedClientConnection<ShardBackendInfo> connectionUp(
- final AbstractClientConnection<ShardBackendInfo> conn, final ShardBackendInfo backend) {
-
- // Step 0: create a new connected connection
- final ConnectedClientConnection<ShardBackendInfo> newConn = new ConnectedClientConnection<>(conn.context(),
- conn.cookie(), backend);
-
- LOG.debug("{}: resolving connection {} to {}", persistenceId(), conn, newConn);
-
+ protected final ConnectionConnectCohort connectionUp(final ConnectedClientConnection<ShardBackendInfo> newConn) {
+ // Step 1: Freeze all AbstractProxyHistory instances pointing to that shard. This indirectly means that no
+ // further TransactionProxies can be created and we can safely traverse maps without risking
+ // missing an entry
final Collection<HistoryReconnectCohort> cohorts = new ArrayList<>();
- try {
- // Step 1: Freeze all AbstractProxyHistory instances pointing to that shard. This indirectly means that no
- // further TransactionProxies can be created and we can safely traverse maps without risking
- // missing an entry
- startReconnect(singleHistory, newConn, cohorts);
- for (ClientLocalHistory h : histories.values()) {
- startReconnect(h, newConn, cohorts);
- }
-
- // Step 2: Collect previous successful requests from the cohorts. We do not want to expose
- // the non-throttling interface to the connection, hence we use a wrapper consumer
- for (HistoryReconnectCohort c : cohorts) {
- c.replaySuccessfulRequests();
- }
-
- // Step 3: Install a forwarder, which will forward requests back to affected cohorts. Any outstanding
- // requests will be immediately sent to it and requests being sent concurrently will get forwarded
- // once they hit the new connection.
- conn.setForwarder(BouncingReconnectForwarder.forCohorts(newConn, cohorts));
- } finally {
- // Step 4: Complete switchover of the connection. The cohorts can resume normal operations.
- for (HistoryReconnectCohort c : cohorts) {
- c.close();
- }
+ startReconnect(singleHistory, newConn, cohorts);
+ for (ClientLocalHistory h : histories.values()) {
+ startReconnect(h, newConn, cohorts);
}
- return newConn;
+ return previousEntries -> {
+ try {
+ // Step 2: Collect previous successful requests from the cohorts. We do not want to expose
+ // the non-throttling interface to the connection, hence we use a wrapper consumer
+ for (HistoryReconnectCohort c : cohorts) {
+ c.replaySuccessfulRequests(previousEntries);
+ }
+
+ // Step 3: Install a forwarder, which will forward requests back to affected cohorts. Any outstanding
+ // requests will be immediately sent to it and requests being sent concurrently will get
+ // forwarded once they hit the new connection.
+ return BouncingReconnectForwarder.forCohorts(newConn, cohorts);
+ } finally {
+ // Step 4: Complete switchover of the connection. The cohorts can resume normal operations.
+ for (HistoryReconnectCohort c : cohorts) {
+ c.close();
+ }
+ }
+ };
}
private static void startReconnect(final AbstractClientHistory history,
import com.google.common.util.concurrent.SettableFuture;
import java.util.ArrayDeque;
import java.util.Deque;
+import java.util.Iterator;
import java.util.concurrent.CountDownLatch;
+import java.util.concurrent.atomic.AtomicIntegerFieldUpdater;
+import java.util.concurrent.atomic.AtomicReferenceFieldUpdater;
import java.util.function.Consumer;
import javax.annotation.Nonnull;
import javax.annotation.Nullable;
import javax.annotation.concurrent.GuardedBy;
import javax.annotation.concurrent.NotThreadSafe;
+import org.opendaylight.controller.cluster.access.client.ConnectionEntry;
import org.opendaylight.controller.cluster.access.commands.TransactionAbortRequest;
import org.opendaylight.controller.cluster.access.commands.TransactionAbortSuccess;
import org.opendaylight.controller.cluster.access.commands.TransactionCanCommitSuccess;
import org.opendaylight.controller.cluster.access.commands.TransactionPreCommitRequest;
import org.opendaylight.controller.cluster.access.commands.TransactionPreCommitSuccess;
import org.opendaylight.controller.cluster.access.commands.TransactionRequest;
-import org.opendaylight.controller.cluster.access.concepts.RequestException;
+import org.opendaylight.controller.cluster.access.concepts.Request;
import org.opendaylight.controller.cluster.access.concepts.RequestFailure;
import org.opendaylight.controller.cluster.access.concepts.Response;
import org.opendaylight.controller.cluster.access.concepts.TransactionIdentifier;
}
}
- private enum SealState {
- /**
- * The user has not sealed the transaction yet.
- */
- OPEN,
- /**
- * The user has sealed the transaction, but has not issued a canCommit().
- */
- SEALED,
- /**
- * The user has sealed the transaction and has issued a canCommit().
- */
- FLUSHED,
+ // Generic state base class. Direct instances are used for fast paths, sub-class is used for successor transitions
+ private static class State {
+ private final String string;
+
+ State(final String string) {
+ this.string = Preconditions.checkNotNull(string);
+ }
+
+ @Override
+ public final String toString() {
+ return string;
+ }
}
- private static final Logger LOG = LoggerFactory.getLogger(AbstractProxyTransaction.class);
+ // State class used when a successor has interfered. Contains coordinator latch, the successor and previous state
+ private static final class SuccessorState extends State {
+ private final CountDownLatch latch = new CountDownLatch(1);
+ private AbstractProxyTransaction successor;
+ private State prevState;
+
+ SuccessorState() {
+ super("successor");
+ }
+
+ // Synchronize with succession process and return the successor
+ AbstractProxyTransaction await() {
+ try {
+ latch.await();
+ } catch (InterruptedException e) {
+ LOG.warn("Interrupted while waiting for latch of {}", successor);
+ throw Throwables.propagate(e);
+ }
+ return successor;
+ }
+
+ void finish() {
+ latch.countDown();
+ }
+
+ State getPrevState() {
+ return prevState;
+ }
+
+ void setPrevState(final State prevState) {
+ Verify.verify(this.prevState == null);
+ this.prevState = Preconditions.checkNotNull(prevState);
+ }
+
+ // To be called from safe contexts, where successor is known to be completed
+ AbstractProxyTransaction getSuccessor() {
+ return Verify.verifyNotNull(successor);
+ }
+ void setSuccessor(final AbstractProxyTransaction successor) {
+ Verify.verify(this.successor == null);
+ this.successor = Preconditions.checkNotNull(successor);
+ }
+ }
+
+ private static final Logger LOG = LoggerFactory.getLogger(AbstractProxyTransaction.class);
+ private static final AtomicIntegerFieldUpdater<AbstractProxyTransaction> SEALED_UPDATER =
+ AtomicIntegerFieldUpdater.newUpdater(AbstractProxyTransaction.class, "sealed");
+ private static final AtomicReferenceFieldUpdater<AbstractProxyTransaction, State> STATE_UPDATER =
+ AtomicReferenceFieldUpdater.newUpdater(AbstractProxyTransaction.class, State.class, "state");
+ private static final State OPEN = new State("open");
+ private static final State SEALED = new State("sealed");
+ private static final State FLUSHED = new State("flushed");
+
+ // Touched from client actor thread only
private final Deque<Object> successfulRequests = new ArrayDeque<>();
private final ProxyHistory parent;
+ // Accessed from user thread only, which may not access this object concurrently
+ private long sequence;
+
/*
* Atomic state-keeping is required to synchronize the process of propagating completed transaction state towards
* the backend -- which may include a successor.
* or timeout, when a successor is injected.
*
* This leaves the problem of needing to completely transferring state just after all queued messages are replayed
- * after a successor was injected, so that it can be properly sealed if we are racing.
+ * after a successor was injected, so that it can be properly sealed if we are racing. Further complication comes
+ * from lock ordering, where the successor injection works with a locked queue and locks proxy objects -- leading
+ * to a potential AB-BA deadlock in case of a naive implementation.
+ *
+ * For tracking user-visible state we use a single volatile int, which is flipped atomically from 0 to 1 exactly
+ * once in {@link AbstractProxyTransaction#seal()}. That keeps common operations fast, as they need to perform
+ * only a single volatile read to assert state correctness.
+ *
+ * For synchronizing client actor (successor-injecting) and user (commit-driving) thread, we keep a separate state
+ * variable. It uses pre-allocated objects for fast paths (i.e. no successor present) and a per-transition object
+ * for slow paths (when successor is injected/present).
*/
- private volatile SealState sealed = SealState.OPEN;
- @GuardedBy("this")
- private AbstractProxyTransaction successor;
- @GuardedBy("this")
- private CountDownLatch successorLatch;
-
- // Accessed from user thread only, which may not access this object concurrently
- private long sequence;
-
+ private volatile int sealed = 0;
+ private volatile State state = OPEN;
AbstractProxyTransaction(final ProxyHistory parent) {
this.parent = Preconditions.checkNotNull(parent);
* Seal this transaction before it is either committed or aborted.
*/
final void seal() {
- final CountDownLatch localLatch;
-
- synchronized (this) {
- checkNotSealed();
- doSeal();
-
- // Fast path: no successor
- if (successor == null) {
- sealed = SealState.SEALED;
- parent.onTransactionSealed(this);
- return;
- }
-
- localLatch = successorLatch;
- }
-
- // Slow path: wait for the latch
- LOG.debug("{} waiting on successor latch", getIdentifier());
- try {
- localLatch.await();
- } catch (InterruptedException e) {
- LOG.warn("{} interrupted while waiting for latch", getIdentifier());
- throw Throwables.propagate(e);
- }
-
- synchronized (this) {
- LOG.debug("{} reacquired lock", getIdentifier());
-
+ // Transition user-visible state first
+ final boolean success = SEALED_UPDATER.compareAndSet(this, 0, 1);
+ Preconditions.checkState(success, "Proxy %s was already sealed", getIdentifier());
+ doSeal();
+ parent.onTransactionSealed(this);
+
+ // Now deal with state transfer, which can occur via successor or a follow-up canCommit() or directCommit().
+ if (!STATE_UPDATER.compareAndSet(this, OPEN, SEALED)) {
+ // Slow path: wait for the successor to complete
+ final AbstractProxyTransaction successor = awaitSuccessor();
+
+ // At this point the successor has completed transition and is possibly visible by the user thread, which is
+ // still stuck here. The successor has not seen final part of our state, nor the fact it is sealed.
+ // Propagate state and seal the successor.
flushState(successor);
successor.seal();
-
- sealed = SealState.FLUSHED;
- parent.onTransactionSealed(this);
}
}
private void checkNotSealed() {
- Preconditions.checkState(sealed == SealState.OPEN, "Transaction %s has already been sealed", getIdentifier());
+ Preconditions.checkState(sealed == 0, "Transaction %s has already been sealed", getIdentifier());
}
- private SealState checkSealed() {
- final SealState local = sealed;
- Preconditions.checkState(local != SealState.OPEN, "Transaction %s has not been sealed yet", getIdentifier());
- return local;
+ private void checkSealed() {
+ Preconditions.checkState(sealed != 0, "Transaction %s has not been sealed yet", getIdentifier());
+ }
+
+ private SuccessorState getSuccessorState() {
+ final State local = state;
+ Verify.verify(local instanceof SuccessorState, "State %s has unexpected class", local);
+ return (SuccessorState) local;
}
final void recordSuccessfulRequest(final @Nonnull TransactionRequest<?> req) {
* @return Future completion
*/
final ListenableFuture<Boolean> directCommit() {
- final CountDownLatch localLatch;
+ checkSealed();
+ // Precludes startReconnect() from interfering with the fast path
synchronized (this) {
- final SealState local = checkSealed();
-
- // Fast path: no successor asserted
- if (successor == null) {
- Verify.verify(local == SealState.SEALED);
-
+ if (STATE_UPDATER.compareAndSet(this, SEALED, FLUSHED)) {
final SettableFuture<Boolean> ret = SettableFuture.create();
sendRequest(Verify.verifyNotNull(commitRequest(false)), t -> {
if (t instanceof TransactionCommitSuccess) {
parent.completeTransaction(this);
});
- sealed = SealState.FLUSHED;
return ret;
}
-
- // We have a successor, take its latch
- localLatch = successorLatch;
- }
-
- // Slow path: we need to wait for the successor to completely propagate
- LOG.debug("{} waiting on successor latch", getIdentifier());
- try {
- localLatch.await();
- } catch (InterruptedException e) {
- LOG.warn("{} interrupted while waiting for latch", getIdentifier());
- throw Throwables.propagate(e);
}
- synchronized (this) {
- LOG.debug("{} reacquired lock", getIdentifier());
-
- final SealState local = sealed;
- Verify.verify(local == SealState.FLUSHED);
-
- return successor.directCommit();
- }
+ // We have had some interference with successor injection, wait for it to complete and defer to the successor.
+ return awaitSuccessor().directCommit();
}
final void canCommit(final VotingFuture<?> ret) {
- final CountDownLatch localLatch;
+ checkSealed();
+ // Precludes startReconnect() from interfering with the fast path
synchronized (this) {
- final SealState local = checkSealed();
-
- // Fast path: no successor asserted
- if (successor == null) {
- Verify.verify(local == SealState.SEALED);
-
+ if (STATE_UPDATER.compareAndSet(this, SEALED, FLUSHED)) {
final TransactionRequest<?> req = Verify.verifyNotNull(commitRequest(true));
+
sendRequest(req, t -> {
if (t instanceof TransactionCanCommitSuccess) {
ret.voteYes();
LOG.debug("Transaction {} canCommit completed", this);
});
- sealed = SealState.FLUSHED;
return;
}
-
- // We have a successor, take its latch
- localLatch = successorLatch;
- }
-
- // Slow path: we need to wait for the successor to completely propagate
- LOG.debug("{} waiting on successor latch", getIdentifier());
- try {
- localLatch.await();
- } catch (InterruptedException e) {
- LOG.warn("{} interrupted while waiting for latch", getIdentifier());
- throw Throwables.propagate(e);
}
- synchronized (this) {
- LOG.debug("{} reacquired lock", getIdentifier());
-
- final SealState local = sealed;
- Verify.verify(local == SealState.FLUSHED);
+ // We have had some interference with successor injection, wait for it to complete and defer to the successor.
+ awaitSuccessor().canCommit(ret);
+ }
- successor.canCommit(ret);
- }
+ private AbstractProxyTransaction awaitSuccessor() {
+ return getSuccessorState().await();
}
final void preCommit(final VotingFuture<?> ret) {
});
}
- void doCommit(final VotingFuture<?> ret) {
+ final void doCommit(final VotingFuture<?> ret) {
checkSealed();
sendRequest(new TransactionDoCommitRequest(getIdentifier(), nextSequence(), localActor()), t -> {
});
}
- final synchronized void startReconnect(final AbstractProxyTransaction successor) {
- Preconditions.checkState(this.successor == null);
- this.successor = Preconditions.checkNotNull(successor);
+ // Called with the connection unlocked
+ final synchronized void startReconnect() {
+ // At this point canCommit/directCommit are blocked, we assert a new successor state, retrieving the previous
+ // state. This method is called with the queue still unlocked.
+ final SuccessorState nextState = new SuccessorState();
+ final State prevState = STATE_UPDATER.getAndSet(this, nextState);
+
+ LOG.debug("Start reconnect of proxy {} previous state {}", this, prevState);
+ Verify.verify(!(prevState instanceof SuccessorState), "Proxy %s duplicate reconnect attempt after %s", this,
+ prevState);
+
+ // We have asserted a slow-path state, seal(), canCommit(), directCommit() are forced to slow paths, which will
+ // wait until we unblock nextState's latch before accessing state. Now we record prevState for later use and we
+ // are done.
+ nextState.setPrevState(prevState);
+ }
+
+ // Called with the connection locked
+ final void replayMessages(final AbstractProxyTransaction successor,
+ final Iterable<ConnectionEntry> enqueuedEntries) {
+ final SuccessorState local = getSuccessorState();
+ local.setSuccessor(successor);
+ // Replay successful requests first
for (Object obj : successfulRequests) {
if (obj instanceof TransactionRequest) {
- LOG.debug("Forwarding request {} to successor {}", obj, successor);
+ LOG.debug("Forwarding successful request {} to successor {}", obj, successor);
successor.handleForwardedRemoteRequest((TransactionRequest<?>) obj, null);
} else {
Verify.verify(obj instanceof IncrementSequence);
LOG.debug("{} replayed {} successful requests", getIdentifier(), successfulRequests.size());
successfulRequests.clear();
+ // Now replay whatever is in the connection
+ final Iterator<ConnectionEntry> it = enqueuedEntries.iterator();
+ while (it.hasNext()) {
+ final ConnectionEntry e = it.next();
+ final Request<?, ?> req = e.getRequest();
+
+ if (getIdentifier().equals(req.getTarget())) {
+ Verify.verify(req instanceof TransactionRequest, "Unhandled request %s", req);
+ LOG.debug("Forwarding queued request{} to successor {}", req, successor);
+ successor.handleForwardedRemoteRequest((TransactionRequest<?>) req, e.getCallback());
+ it.remove();
+ }
+ }
+
/*
- * Before releasing the lock we need to make sure that a call to seal() blocks until we have completed
- * finishConnect().
+ * Check the state at which we have started the reconnect attempt. State transitions triggered while we were
+ * reconnecting have been forced to slow paths, which will be unlocked once we unblock the state latch
+ * at the end of this method.
*/
- successorLatch = new CountDownLatch(1);
- }
-
- final synchronized void finishReconnect() {
- Preconditions.checkState(successorLatch != null);
-
- if (sealed == SealState.SEALED) {
- /*
- * If this proxy is in the 'sealed, have not sent canCommit' state. If so, we need to forward current
- * leftover state to the successor now.
- */
+ final State prevState = local.getPrevState();
+ if (SEALED.equals(prevState)) {
+ LOG.debug("Proxy {} reconnected while being sealed, propagating state to successor {}", this, successor);
flushState(successor);
successor.seal();
- sealed = SealState.FLUSHED;
}
+ }
+
+ // Called with the connection locked
+ final void finishReconnect() {
+ final SuccessorState local = getSuccessorState();
+ LOG.debug("Finishing reconnect of proxy {}", this);
- // All done, release the latch, unblocking seal() and canCommit()
- successorLatch.countDown();
+ // All done, release the latch, unblocking seal() and canCommit() slow paths
+ local.finish();
}
/**
*
* @param request Request to be forwarded
* @param callback Original callback
- * @throws RequestException when the request is unhandled by the successor
*/
- final synchronized void replayRequest(final TransactionRequest<?> request,
- final Consumer<Response<?, ?>> callback) {
- Preconditions.checkState(successor != null, "%s does not have a successor set", this);
+ final void replayRequest(final TransactionRequest<?> request, final Consumer<Response<?, ?>> callback) {
+ final AbstractProxyTransaction successor = getSuccessorState().getSuccessor();
if (successor instanceof LocalProxyTransaction) {
forwardToLocal((LocalProxyTransaction)successor, request, callback);
@Override
- protected void forwardEntry(final ConnectionEntry entry) {
+ protected void forwardEntry(final ConnectionEntry entry, final long now) {
final Request<? , ?> request = entry.getRequest();
final LocalHistoryIdentifier historyId;
*/
package org.opendaylight.controller.cluster.databroker.actors.dds;
+import org.opendaylight.controller.cluster.access.client.ConnectionEntry;
+
/**
* Interface exposed by {@link AbstractClientHistory} to {@link DistributedDataStoreClientBehavior} for the sole
* purpose of performing a connection switchover.
abstract class HistoryReconnectCohort implements AutoCloseable {
abstract ProxyReconnectCohort getProxy();
- abstract void replaySuccessfulRequests();
+ abstract void replaySuccessfulRequests(Iterable<ConnectionEntry> previousEntries);
@Override
public abstract void close();
import javax.annotation.concurrent.GuardedBy;
import org.opendaylight.controller.cluster.access.client.AbstractClientConnection;
import org.opendaylight.controller.cluster.access.client.ConnectedClientConnection;
+import org.opendaylight.controller.cluster.access.client.ConnectionEntry;
+import org.opendaylight.controller.cluster.access.commands.CreateLocalHistoryRequest;
import org.opendaylight.controller.cluster.access.commands.LocalHistoryRequest;
import org.opendaylight.controller.cluster.access.commands.TransactionRequest;
import org.opendaylight.controller.cluster.access.concepts.LocalHistoryIdentifier;
@GuardedBy("lock")
@Override
- void replaySuccessfulRequests() {
+ void replaySuccessfulRequests(final Iterable<ConnectionEntry> previousEntries) {
+ // First look for our Create message
+ for (ConnectionEntry e : previousEntries) {
+ final Request<?, ?> req = e.getRequest();
+ if (identifier.equals(req.getTarget())) {
+ Verify.verify(req instanceof LocalHistoryRequest);
+ if (req instanceof CreateLocalHistoryRequest) {
+ successor.connection.sendRequest(req, e.getCallback());
+ break;
+ }
+ }
+ }
+
for (AbstractProxyTransaction t : proxies.values()) {
LOG.debug("{} creating successor transaction proxy for {}", identifier, t);
final AbstractProxyTransaction newProxy = successor.createTransactionProxy(t.getIdentifier());
LOG.debug("{} created successor transaction proxy {}", identifier, newProxy);
- t.startReconnect(newProxy);
+ t.replayMessages(newProxy, previousEntries);
+ }
+
+ // Now look for any finalizing messages
+ for (ConnectionEntry e : previousEntries) {
+ final Request<?, ?> req = e.getRequest();
+ if (identifier.equals(req.getTarget())) {
+ Verify.verify(req instanceof LocalHistoryRequest);
+ successor.connection.sendRequest(req, e.getCallback());
+ }
}
}
successor = createSuccessor(newConnection);
LOG.debug("History {} instantiated successor {}", this, successor);
+
+ for (AbstractProxyTransaction t : proxies.values()) {
+ t.startReconnect();
+ }
+
return new ReconnectCohort();
}
import java.util.function.BiConsumer;
import java.util.function.Consumer;
+import org.opendaylight.controller.cluster.access.client.ConnectionEntry;
import org.opendaylight.controller.cluster.access.concepts.LocalHistoryIdentifier;
import org.opendaylight.controller.cluster.access.concepts.Request;
import org.opendaylight.controller.cluster.access.concepts.RequestException;
abstract class ProxyReconnectCohort implements Identifiable<LocalHistoryIdentifier> {
- abstract void replaySuccessfulRequests();
+ abstract void replaySuccessfulRequests(Iterable<ConnectionEntry> previousEntries);
abstract ProxyHistory finishReconnect();
Code Review / controller.git / commitdiff