import com.google.common.collect.ImmutableMap.Builder;
import com.google.common.collect.Iterables;
import com.google.common.primitives.UnsignedLong;
+import com.google.common.util.concurrent.FutureCallback;
import edu.umd.cs.findbugs.annotations.SuppressFBWarnings;
import java.io.File;
import java.io.IOException;
import java.util.ArrayDeque;
import java.util.ArrayList;
import java.util.Collection;
+import java.util.Collections;
+import java.util.Deque;
import java.util.HashMap;
import java.util.Iterator;
import java.util.Map;
import java.util.Map.Entry;
import java.util.Queue;
-import java.util.concurrent.ExecutionException;
+import java.util.SortedSet;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.TimeoutException;
import java.util.function.Consumer;
+import java.util.function.Function;
import java.util.function.UnaryOperator;
import javax.annotation.Nonnull;
import javax.annotation.Nullable;
import org.opendaylight.controller.cluster.datastore.utils.DataTreeModificationOutput;
import org.opendaylight.controller.cluster.datastore.utils.PruningDataTreeModification;
import org.opendaylight.controller.cluster.raft.protobuff.client.messages.Payload;
-import org.opendaylight.controller.md.sal.common.api.data.AsyncDataBroker.DataChangeScope;
-import org.opendaylight.controller.md.sal.common.api.data.AsyncDataChangeListener;
-import org.opendaylight.controller.md.sal.common.api.data.OptimisticLockFailedException;
-import org.opendaylight.controller.md.sal.common.api.data.TransactionCommitFailedException;
-import org.opendaylight.controller.md.sal.dom.api.DOMDataTreeChangeListener;
+import org.opendaylight.mdsal.common.api.OptimisticLockFailedException;
+import org.opendaylight.mdsal.common.api.TransactionCommitFailedException;
+import org.opendaylight.mdsal.dom.api.DOMDataTreeChangeListener;
import org.opendaylight.yangtools.concepts.Identifier;
import org.opendaylight.yangtools.concepts.ListenerRegistration;
import org.opendaylight.yangtools.yang.data.api.YangInstanceIdentifier;
import org.opendaylight.yangtools.yang.data.api.schema.NormalizedNode;
import org.opendaylight.yangtools.yang.data.api.schema.tree.ConflictingModificationAppliedException;
+import org.opendaylight.yangtools.yang.data.api.schema.tree.DataTree;
import org.opendaylight.yangtools.yang.data.api.schema.tree.DataTreeCandidate;
import org.opendaylight.yangtools.yang.data.api.schema.tree.DataTreeCandidateTip;
import org.opendaylight.yangtools.yang.data.api.schema.tree.DataTreeCandidates;
+import org.opendaylight.yangtools.yang.data.api.schema.tree.DataTreeConfiguration;
import org.opendaylight.yangtools.yang.data.api.schema.tree.DataTreeModification;
import org.opendaylight.yangtools.yang.data.api.schema.tree.DataTreeSnapshot;
import org.opendaylight.yangtools.yang.data.api.schema.tree.DataTreeTip;
import org.opendaylight.yangtools.yang.data.api.schema.tree.DataValidationFailedException;
-import org.opendaylight.yangtools.yang.data.api.schema.tree.TipProducingDataTree;
-import org.opendaylight.yangtools.yang.data.api.schema.tree.TipProducingDataTreeTip;
import org.opendaylight.yangtools.yang.data.api.schema.tree.TreeType;
import org.opendaylight.yangtools.yang.data.impl.schema.tree.InMemoryDataTreeFactory;
import org.opendaylight.yangtools.yang.model.api.SchemaContext;
this.cohort = Preconditions.checkNotNull(cohort);
lastAccess = now;
}
+
+ @Override
+ public String toString() {
+ return "CommitEntry [tx=" + cohort.getIdentifier() + ", state=" + cohort.getState() + "]";
+ }
}
private static final Timeout COMMIT_STEP_TIMEOUT = new Timeout(Duration.create(5, TimeUnit.SECONDS));
private final Map<LocalHistoryIdentifier, ShardDataTreeTransactionChain> transactionChains = new HashMap<>();
private final DataTreeCohortActorRegistry cohortRegistry = new DataTreeCohortActorRegistry();
- private final Queue<CommitEntry> pendingTransactions = new ArrayDeque<>();
+ private final Deque<CommitEntry> pendingTransactions = new ArrayDeque<>();
private final Queue<CommitEntry> pendingCommits = new ArrayDeque<>();
private final Queue<CommitEntry> pendingFinishCommits = new ArrayDeque<>();
private final Map<Payload, Runnable> replicationCallbacks = new HashMap<>();
private final ShardDataTreeChangeListenerPublisher treeChangeListenerPublisher;
- private final ShardDataChangeListenerPublisher dataChangeListenerPublisher;
private final Collection<ShardDataTreeMetadata<?>> metadata;
- private final TipProducingDataTree dataTree;
+ private final DataTree dataTree;
private final String logContext;
private final Shard shard;
private Runnable runOnPendingTransactionsComplete;
/**
* Optimistic {@link DataTreeCandidate} preparation. Since our DataTree implementation is a
- * {@link TipProducingDataTree}, each {@link DataTreeCandidate} is also a {@link DataTreeTip}, e.g. another
+ * {@link DataTree}, each {@link DataTreeCandidate} is also a {@link DataTreeTip}, e.g. another
* candidate can be prepared on top of it. They still need to be committed in sequence. Here we track the current
* tip of the data tree, which is the last DataTreeCandidate we have in flight, or the DataTree itself.
*/
- private TipProducingDataTreeTip tip;
+ private DataTreeTip tip;
private SchemaContext schemaContext;
private int currentTransactionBatch;
- ShardDataTree(final Shard shard, final SchemaContext schemaContext, final TipProducingDataTree dataTree,
+ ShardDataTree(final Shard shard, final SchemaContext schemaContext, final DataTree dataTree,
final ShardDataTreeChangeListenerPublisher treeChangeListenerPublisher,
- final ShardDataChangeListenerPublisher dataChangeListenerPublisher, final String logContext,
+ final String logContext,
final ShardDataTreeMetadata<?>... metadata) {
this.dataTree = Preconditions.checkNotNull(dataTree);
updateSchemaContext(schemaContext);
this.shard = Preconditions.checkNotNull(shard);
this.treeChangeListenerPublisher = Preconditions.checkNotNull(treeChangeListenerPublisher);
- this.dataChangeListenerPublisher = Preconditions.checkNotNull(dataChangeListenerPublisher);
this.logContext = Preconditions.checkNotNull(logContext);
this.metadata = ImmutableList.copyOf(metadata);
tip = dataTree;
ShardDataTree(final Shard shard, final SchemaContext schemaContext, final TreeType treeType,
final YangInstanceIdentifier root,
final ShardDataTreeChangeListenerPublisher treeChangeListenerPublisher,
- final ShardDataChangeListenerPublisher dataChangeListenerPublisher, final String logContext,
+ final String logContext,
final ShardDataTreeMetadata<?>... metadata) {
- this(shard, schemaContext, InMemoryDataTreeFactory.getInstance().create(treeType, root),
- treeChangeListenerPublisher, dataChangeListenerPublisher, logContext, metadata);
+ this(shard, schemaContext, createDataTree(treeType, root), treeChangeListenerPublisher, logContext, metadata);
+ }
+
+ private static DataTree createDataTree(final TreeType treeType, final YangInstanceIdentifier root) {
+ final DataTreeConfiguration baseConfig = DataTreeConfiguration.getDefault(treeType);
+ return new InMemoryDataTreeFactory().create(new DataTreeConfiguration.Builder(baseConfig.getTreeType())
+ .setMandatoryNodesValidation(baseConfig.isMandatoryNodesValidationEnabled())
+ .setUniqueIndexes(baseConfig.isUniqueIndexEnabled())
+ .setRootPath(root)
+ .build());
}
@VisibleForTesting
public ShardDataTree(final Shard shard, final SchemaContext schemaContext, final TreeType treeType) {
this(shard, schemaContext, treeType, YangInstanceIdentifier.EMPTY,
- new DefaultShardDataTreeChangeListenerPublisher(""),
- new DefaultShardDataChangeListenerPublisher(""), "");
+ new DefaultShardDataTreeChangeListenerPublisher(""), "");
}
final String logContext() {
return shard.ticker().read();
}
- public TipProducingDataTree getDataTree() {
+ public DataTree getDataTree() {
return dataTree;
}
* @param snapshot Snapshot that needs to be applied
* @throws DataValidationFailedException when the snapshot fails to apply
*/
- void applyRecoverySnapshot(final @Nonnull ShardDataTreeSnapshot snapshot) throws DataValidationFailedException {
+ void applyRecoverySnapshot(@Nonnull final ShardDataTreeSnapshot snapshot) throws DataValidationFailedException {
applySnapshot(snapshot, this::wrapWithPruning);
}
@SuppressWarnings("checkstyle:IllegalCatch")
- private void applyRecoveryCandidate(final DataTreeCandidate candidate) throws DataValidationFailedException {
+ private void applyRecoveryCandidate(final DataTreeCandidate candidate) {
final PruningDataTreeModification mod = wrapWithPruning(dataTree.takeSnapshot().newModification());
DataTreeCandidates.applyToModification(mod, candidate);
mod.ready();
* @throws IOException when the snapshot fails to deserialize
* @throws DataValidationFailedException when the snapshot fails to apply
*/
- void applyRecoveryPayload(final @Nonnull Payload payload) throws IOException, DataValidationFailedException {
+ void applyRecoveryPayload(@Nonnull final Payload payload) throws IOException {
if (payload instanceof CommitTransactionPayload) {
final Entry<TransactionIdentifier, DataTreeCandidate> e =
((CommitTransactionPayload) payload).getCandidate();
@VisibleForTesting
public void notifyListeners(final DataTreeCandidate candidate) {
treeChangeListenerPublisher.publishChanges(candidate);
- dataChangeListenerPublisher.publishChanges(candidate);
}
/**
replicatePayload(id, PurgeLocalHistoryPayload.create(id), callback);
}
- void registerDataChangeListener(final YangInstanceIdentifier path,
- final AsyncDataChangeListener<YangInstanceIdentifier, NormalizedNode<?, ?>> listener,
- final DataChangeScope scope, final Optional<DataTreeCandidate> initialState,
- final Consumer<ListenerRegistration<AsyncDataChangeListener<YangInstanceIdentifier, NormalizedNode<?, ?>>>>
- onRegistration) {
- dataChangeListenerPublisher.registerDataChangeListener(path, listener, scope, initialState, onRegistration);
- }
-
Optional<DataTreeCandidate> readCurrentData() {
- final Optional<NormalizedNode<?, ?>> currentState =
+ final java.util.Optional<NormalizedNode<?, ?>> currentState =
dataTree.takeSnapshot().readNode(YangInstanceIdentifier.EMPTY);
return currentState.isPresent() ? Optional.of(DataTreeCandidates.fromNormalizedNode(
YangInstanceIdentifier.EMPTY, currentState.get())) : Optional.<DataTreeCandidate>absent();
}
@Override
- ShardDataTreeCohort finishTransaction(final ReadWriteShardDataTreeTransaction transaction) {
+ ShardDataTreeCohort finishTransaction(final ReadWriteShardDataTreeTransaction transaction,
+ final java.util.Optional<SortedSet<String>> participatingShardNames) {
final DataTreeModification snapshot = transaction.getSnapshot();
snapshot.ready();
- return createReadyCohort(transaction.getIdentifier(), snapshot);
+ return createReadyCohort(transaction.getIdentifier(), snapshot, participatingShardNames);
}
void purgeTransaction(final TransactionIdentifier id, final Runnable callback) {
}
public Optional<NormalizedNode<?, ?>> readNode(final YangInstanceIdentifier path) {
- return dataTree.takeSnapshot().readNode(path);
+ return Optional.fromJavaUtil(dataTree.takeSnapshot().readNode(path));
}
DataTreeSnapshot takeSnapshot() {
} catch (ConflictingModificationAppliedException e) {
LOG.warn("{}: Store Tx {}: Conflicting modification for path {}.", logContext, cohort.getIdentifier(),
e.getPath());
- cause = new OptimisticLockFailedException("Optimistic lock failed.", e);
+ cause = new OptimisticLockFailedException("Optimistic lock failed for path " + e.getPath(), e);
} catch (DataValidationFailedException e) {
LOG.warn("{}: Store Tx {}: Data validation failed for path {}.", logContext, cohort.getIdentifier(),
e.getPath(), e);
// precondition log, it should allow us to understand what went on.
LOG.debug("{}: Store Tx {}: modifications: {} tree: {}", cohort.getIdentifier(), modification,
dataTree);
- cause = new TransactionCommitFailedException("Data did not pass validation.", e);
+ cause = new TransactionCommitFailedException("Data did not pass validation for path " + e.getPath(), e);
} catch (Exception e) {
LOG.warn("{}: Unexpected failure in validation phase", logContext, e);
cause = e;
return;
}
if (!cohort.equals(head.cohort)) {
- LOG.debug("{}: Transaction {} scheduled for canCommit step", logContext, cohort.getIdentifier());
- return;
+ // The tx isn't at the head of the queue so we can't start canCommit at this point. Here we check if this
+ // tx should be moved ahead of other tx's in the READY state in the pendingTransactions queue. If this tx
+ // has other participating shards, it could deadlock with other tx's accessing the same shards
+ // depending on the order the tx's are readied on each shard
+ // (see https://jira.opendaylight.org/browse/CONTROLLER-1836). Therefore, if the preceding participating
+ // shard names for a preceding pending tx, call it A, in the queue matches that of this tx, then this tx
+ // is allowed to be moved ahead of tx A in the queue so it is processed first to avoid potential deadlock
+ // if tx A is behind this tx in the pendingTransactions queue for a preceding shard. In other words, since
+ // canCommmit for this tx was requested before tx A, honor that request. If this tx is moved to the head of
+ // the queue as a result, then proceed with canCommit.
+
+ Collection<String> precedingShardNames = extractPrecedingShardNames(cohort.getParticipatingShardNames());
+ if (precedingShardNames.isEmpty()) {
+ LOG.debug("{}: Tx {} is scheduled for canCommit step", logContext, cohort.getIdentifier());
+ return;
+ }
+
+ LOG.debug("{}: Evaluating tx {} for canCommit - preceding participating shard names {}",
+ logContext, cohort.getIdentifier(), precedingShardNames);
+ final Iterator<CommitEntry> iter = pendingTransactions.iterator();
+ int index = -1;
+ int moveToIndex = -1;
+ while (iter.hasNext()) {
+ final CommitEntry entry = iter.next();
+ ++index;
+
+ if (cohort.equals(entry.cohort)) {
+ if (moveToIndex < 0) {
+ LOG.debug("{}: Not moving tx {} - cannot proceed with canCommit",
+ logContext, cohort.getIdentifier());
+ return;
+ }
+
+ LOG.debug("{}: Moving {} to index {} in the pendingTransactions queue",
+ logContext, cohort.getIdentifier(), moveToIndex);
+ iter.remove();
+ insertEntry(pendingTransactions, entry, moveToIndex);
+
+ if (!cohort.equals(pendingTransactions.peek().cohort)) {
+ LOG.debug("{}: Tx {} is not at the head of the queue - cannot proceed with canCommit",
+ logContext, cohort.getIdentifier());
+ return;
+ }
+
+ LOG.debug("{}: Tx {} is now at the head of the queue - proceeding with canCommit",
+ logContext, cohort.getIdentifier());
+ break;
+ }
+
+ if (entry.cohort.getState() != State.READY) {
+ LOG.debug("{}: Skipping pending transaction {} in state {}",
+ logContext, entry.cohort.getIdentifier(), entry.cohort.getState());
+ continue;
+ }
+
+ final Collection<String> pendingPrecedingShardNames = extractPrecedingShardNames(
+ entry.cohort.getParticipatingShardNames());
+
+ if (precedingShardNames.equals(pendingPrecedingShardNames)) {
+ if (moveToIndex < 0) {
+ LOG.debug("{}: Preceding shard names {} for pending tx {} match - saving moveToIndex {}",
+ logContext, pendingPrecedingShardNames, entry.cohort.getIdentifier(), index);
+ moveToIndex = index;
+ } else {
+ LOG.debug(
+ "{}: Preceding shard names {} for pending tx {} match but moveToIndex already set to {}",
+ logContext, pendingPrecedingShardNames, entry.cohort.getIdentifier(), moveToIndex);
+ }
+ } else {
+ LOG.debug("{}: Preceding shard names {} for pending tx {} differ - skipping",
+ logContext, pendingPrecedingShardNames, entry.cohort.getIdentifier());
+ }
+ }
}
processNextPendingTransaction();
}
- private void failPreCommit(final Exception cause) {
+ private void insertEntry(Deque<CommitEntry> queue, CommitEntry entry, int atIndex) {
+ if (atIndex == 0) {
+ queue.addFirst(entry);
+ return;
+ }
+
+ LOG.trace("Inserting into Deque at index {}", atIndex);
+
+ Deque<CommitEntry> tempStack = new ArrayDeque<>(atIndex);
+ for (int i = 0; i < atIndex; i++) {
+ tempStack.push(queue.poll());
+ }
+
+ queue.addFirst(entry);
+
+ tempStack.forEach(queue::addFirst);
+ }
+
+ private Collection<String> extractPrecedingShardNames(
+ java.util.Optional<SortedSet<String>> participatingShardNames) {
+ return participatingShardNames.map((Function<SortedSet<String>, Collection<String>>)
+ set -> set.headSet(shard.getShardName())).orElse(Collections.<String>emptyList());
+ }
+
+ private void failPreCommit(final Throwable cause) {
shard.getShardMBean().incrementFailedTransactionsCount();
pendingTransactions.poll().cohort.failedPreCommit(cause);
processNextPendingTransaction();
final DataTreeCandidateTip candidate;
try {
candidate = tip.prepare(cohort.getDataTreeModification());
- cohort.userPreCommit(candidate);
- } catch (ExecutionException | TimeoutException | RuntimeException e) {
+ } catch (RuntimeException e) {
failPreCommit(e);
return;
}
- // Set the tip of the data tree.
- tip = Verify.verifyNotNull(candidate);
+ cohort.userPreCommit(candidate, new FutureCallback<Void>() {
+ @Override
+ public void onSuccess(final Void noop) {
+ // Set the tip of the data tree.
+ tip = Verify.verifyNotNull(candidate);
+
+ entry.lastAccess = readTime();
- entry.lastAccess = readTime();
+ pendingTransactions.remove();
+ pendingCommits.add(entry);
- pendingTransactions.remove();
- pendingCommits.add(entry);
+ LOG.debug("{}: Transaction {} prepared", logContext, current.getIdentifier());
- LOG.debug("{}: Transaction {} prepared", logContext, current.getIdentifier());
+ cohort.successfulPreCommit(candidate);
- cohort.successfulPreCommit(candidate);
+ processNextPendingTransaction();
+ }
- processNextPendingTransaction();
+ @Override
+ public void onFailure(final Throwable failure) {
+ failPreCommit(failure);
+ }
+ });
}
private void failCommit(final Exception cause) {
shard.getShardMBean().setLastCommittedTransactionTime(System.currentTimeMillis());
// FIXME: propagate journal index
- pendingFinishCommits.poll().cohort.successfulCommit(UnsignedLong.ZERO);
-
- LOG.trace("{}: Transaction {} committed, proceeding to notify", logContext, txId);
- notifyListeners(candidate);
+ pendingFinishCommits.poll().cohort.successfulCommit(UnsignedLong.ZERO, () -> {
+ LOG.trace("{}: Transaction {} committed, proceeding to notify", logContext, txId);
+ notifyListeners(candidate);
- processNextPending();
+ processNextPending();
+ });
}
void startCommit(final SimpleShardDataTreeCohort cohort, final DataTreeCandidate candidate) {
}
@Override
- ShardDataTreeCohort createReadyCohort(final TransactionIdentifier txId, final DataTreeModification mod) {
+ ShardDataTreeCohort createReadyCohort(final TransactionIdentifier txId, final DataTreeModification mod,
+ final java.util.Optional<SortedSet<String>> participatingShardNames) {
SimpleShardDataTreeCohort cohort = new SimpleShardDataTreeCohort(this, mod, txId,
- cohortRegistry.createCohort(schemaContext, txId, COMMIT_STEP_TIMEOUT));
+ cohortRegistry.createCohort(schemaContext, txId, shard::executeInSelf,
+ COMMIT_STEP_TIMEOUT), participatingShardNames);
pendingTransactions.add(new CommitEntry(cohort, readTime()));
return cohort;
}
// Exposed for ShardCommitCoordinator so it does not have deal with local histories (it does not care), this mimics
// the newReadWriteTransaction()
- ShardDataTreeCohort newReadyCohort(final TransactionIdentifier txId, final DataTreeModification mod) {
+ ShardDataTreeCohort newReadyCohort(final TransactionIdentifier txId, final DataTreeModification mod,
+ final java.util.Optional<SortedSet<String>> participatingShardNames) {
if (txId.getHistoryId().getHistoryId() == 0) {
- return createReadyCohort(txId, mod);
+ return createReadyCohort(txId, mod, participatingShardNames);
}
- return ensureTransactionChain(txId.getHistoryId(), null).createReadyCohort(txId, mod);
+ return ensureTransactionChain(txId.getHistoryId(), null).createReadyCohort(txId, mod, participatingShardNames);
}
@SuppressFBWarnings(value = "DB_DUPLICATE_SWITCH_CLAUSES", justification = "See inline comments below.")
- void checkForExpiredTransactions(final long transactionCommitTimeoutMillis) {
+ void checkForExpiredTransactions(final long transactionCommitTimeoutMillis,
+ final Function<SimpleShardDataTreeCohort, Optional<Long>> accessTimeUpdater) {
final long timeout = TimeUnit.MILLISECONDS.toNanos(transactionCommitTimeoutMillis);
final long now = readTime();
final Queue<CommitEntry> currentQueue = !pendingFinishCommits.isEmpty() ? pendingFinishCommits :
!pendingCommits.isEmpty() ? pendingCommits : pendingTransactions;
final CommitEntry currentTx = currentQueue.peek();
- if (currentTx != null && currentTx.lastAccess + timeout < now) {
- final State state = currentTx.cohort.getState();
- LOG.warn("{}: Current transaction {} has timed out after {} ms in state {}", logContext,
- currentTx.cohort.getIdentifier(), transactionCommitTimeoutMillis, state);
- boolean processNext = true;
- final TimeoutException cohortFailure = new TimeoutException("Backend timeout in state " + state + " after "
- + transactionCommitTimeoutMillis + "ms");
-
- switch (state) {
- case CAN_COMMIT_PENDING:
- currentQueue.remove().cohort.failedCanCommit(cohortFailure);
- break;
- case CAN_COMMIT_COMPLETE:
- // The suppression of the FindBugs "DB_DUPLICATE_SWITCH_CLAUSES" warning pertains to this clause
- // whose code is duplicated with PRE_COMMIT_COMPLETE. The clauses aren't combined in case the code
- // in PRE_COMMIT_COMPLETE is changed.
- currentQueue.remove().cohort.reportFailure(cohortFailure);
- break;
- case PRE_COMMIT_PENDING:
- currentQueue.remove().cohort.failedPreCommit(cohortFailure);
- break;
- case PRE_COMMIT_COMPLETE:
- // FIXME: this is a legacy behavior problem. Three-phase commit protocol specifies that after we
- // are ready we should commit the transaction, not abort it. Our current software stack does
- // not allow us to do that consistently, because we persist at the time of commit, hence
- // we can end up in a state where we have pre-committed a transaction, then a leader failover
- // occurred ... the new leader does not see the pre-committed transaction and does not have
- // a running timer. To fix this we really need two persistence events.
- //
- // The first one, done at pre-commit time will hold the transaction payload. When consensus
- // is reached, we exit the pre-commit phase and start the pre-commit timer. Followers do not
- // apply the state in this event.
- //
- // The second one, done at commit (or abort) time holds only the transaction identifier and
- // signals to followers that the state should (or should not) be applied.
- //
- // In order to make the pre-commit timer working across failovers, though, we need
- // a per-shard cluster-wide monotonic time, so a follower becoming the leader can accurately
- // restart the timer.
- currentQueue.remove().cohort.reportFailure(cohortFailure);
- break;
- case COMMIT_PENDING:
- LOG.warn("{}: Transaction {} is still committing, cannot abort", logContext,
- currentTx.cohort.getIdentifier());
- currentTx.lastAccess = now;
- processNext = false;
- return;
- case READY:
- currentQueue.remove().cohort.reportFailure(cohortFailure);
- break;
- case ABORTED:
- case COMMITTED:
- case FAILED:
- default:
- currentQueue.remove();
+ if (currentTx == null) {
+ // Empty queue, no-op
+ return;
+ }
+
+ long delta = now - currentTx.lastAccess;
+ if (delta < timeout) {
+ // Not expired yet, bail
+ return;
+ }
+
+ final Optional<Long> updateOpt = accessTimeUpdater.apply(currentTx.cohort);
+ if (updateOpt.isPresent()) {
+ final long newAccess = updateOpt.get().longValue();
+ final long newDelta = now - newAccess;
+ if (newDelta < delta) {
+ LOG.debug("{}: Updated current transaction {} access time", logContext,
+ currentTx.cohort.getIdentifier());
+ currentTx.lastAccess = newAccess;
+ delta = newDelta;
}
- if (processNext) {
- processNextPending();
+ if (delta < timeout) {
+ // Not expired yet, bail
+ return;
}
}
+
+ final long deltaMillis = TimeUnit.NANOSECONDS.toMillis(delta);
+ final State state = currentTx.cohort.getState();
+
+ LOG.warn("{}: Current transaction {} has timed out after {} ms in state {}", logContext,
+ currentTx.cohort.getIdentifier(), deltaMillis, state);
+ boolean processNext = true;
+ final TimeoutException cohortFailure = new TimeoutException("Backend timeout in state " + state + " after "
+ + deltaMillis + "ms");
+
+ switch (state) {
+ case CAN_COMMIT_PENDING:
+ currentQueue.remove().cohort.failedCanCommit(cohortFailure);
+ break;
+ case CAN_COMMIT_COMPLETE:
+ // The suppression of the FindBugs "DB_DUPLICATE_SWITCH_CLAUSES" warning pertains to this clause
+ // whose code is duplicated with PRE_COMMIT_COMPLETE. The clauses aren't combined in case the code
+ // in PRE_COMMIT_COMPLETE is changed.
+ currentQueue.remove().cohort.reportFailure(cohortFailure);
+ break;
+ case PRE_COMMIT_PENDING:
+ currentQueue.remove().cohort.failedPreCommit(cohortFailure);
+ break;
+ case PRE_COMMIT_COMPLETE:
+ // FIXME: this is a legacy behavior problem. Three-phase commit protocol specifies that after we
+ // are ready we should commit the transaction, not abort it. Our current software stack does
+ // not allow us to do that consistently, because we persist at the time of commit, hence
+ // we can end up in a state where we have pre-committed a transaction, then a leader failover
+ // occurred ... the new leader does not see the pre-committed transaction and does not have
+ // a running timer. To fix this we really need two persistence events.
+ //
+ // The first one, done at pre-commit time will hold the transaction payload. When consensus
+ // is reached, we exit the pre-commit phase and start the pre-commit timer. Followers do not
+ // apply the state in this event.
+ //
+ // The second one, done at commit (or abort) time holds only the transaction identifier and
+ // signals to followers that the state should (or should not) be applied.
+ //
+ // In order to make the pre-commit timer working across failovers, though, we need
+ // a per-shard cluster-wide monotonic time, so a follower becoming the leader can accurately
+ // restart the timer.
+ currentQueue.remove().cohort.reportFailure(cohortFailure);
+ break;
+ case COMMIT_PENDING:
+ LOG.warn("{}: Transaction {} is still committing, cannot abort", logContext,
+ currentTx.cohort.getIdentifier());
+ currentTx.lastAccess = now;
+ processNext = false;
+ return;
+ case READY:
+ currentQueue.remove().cohort.reportFailure(cohortFailure);
+ break;
+ case ABORTED:
+ case COMMITTED:
+ case FAILED:
+ default:
+ currentQueue.remove();
+ }
+
+ if (processNext) {
+ processNextPending();
+ }
}
boolean startAbort(final SimpleShardDataTreeCohort cohort) {
return false;
}
- TipProducingDataTreeTip newTip = MoreObjects.firstNonNull(first.cohort.getCandidate(), dataTree);
+ DataTreeTip newTip = MoreObjects.firstNonNull(first.cohort.getCandidate(), dataTree);
while (it.hasNext()) {
final CommitEntry e = it.next();
if (cohort.equals(e.cohort)) {
}
@SuppressWarnings("checkstyle:IllegalCatch")
- private void rebaseTransactions(final Iterator<CommitEntry> iter, @Nonnull final TipProducingDataTreeTip newTip) {
+ private void rebaseTransactions(final Iterator<CommitEntry> iter, @Nonnull final DataTreeTip newTip) {
tip = Preconditions.checkNotNull(newTip);
while (iter.hasNext()) {
final SimpleShardDataTreeCohort cohort = iter.next().cohort;
try {
tip.validate(cohort.getDataTreeModification());
DataTreeCandidateTip candidate = tip.prepare(cohort.getDataTreeModification());
- cohort.userPreCommit(candidate);
cohort.setNewCandidate(candidate);
tip = candidate;
- } catch (ExecutionException | TimeoutException | RuntimeException | DataValidationFailedException e) {
+ } catch (RuntimeException | DataValidationFailedException e) {
LOG.debug("{}: Failed to reprepare queued transaction {}", logContext, cohort.getIdentifier(), e);
cohort.reportFailure(e);
}
Code Review / controller.git / blobdiff