pnowojski commented on a change in pull request #11899: URL: https://github.com/apache/flink/pull/11899#discussion_r419410190
########## File path: flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/CheckpointRequestDecider.java ########## @@ -0,0 +1,156 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.flink.runtime.checkpoint; + +import org.apache.flink.annotation.VisibleForTesting; +import org.apache.flink.runtime.checkpoint.CheckpointCoordinator.CheckpointTriggerRequest; +import org.apache.flink.runtime.util.clock.Clock; + +import java.util.Optional; +import java.util.PriorityQueue; +import java.util.function.Consumer; +import java.util.function.Function; +import java.util.function.Supplier; + +import static org.apache.flink.runtime.checkpoint.CheckpointFailureReason.MINIMUM_TIME_BETWEEN_CHECKPOINTS; +import static org.apache.flink.runtime.checkpoint.CheckpointFailureReason.TOO_MANY_CONCURRENT_CHECKPOINTS; + +@SuppressWarnings("ConstantConditions") +class CheckpointRequestDecider { + private static final int MAX_QUEUED_REQUESTS = 1000; + + private final int maxConcurrentCheckpointAttempts; + private final Consumer<Long> rescheduleTrigger; + private final Clock clock; + private final long minPauseBetweenCheckpoints; + private final Supplier<Integer> pendingCheckpointsSizeSupplier; + private final Function<Supplier<Optional<CheckpointTriggerRequest>>, Optional<CheckpointTriggerRequest>> sync; + private final PriorityQueue<CheckpointTriggerRequest> triggerRequestQueue = new PriorityQueue<>((r1, r2) -> { + if (r1.props.isSavepoint() != r2.props.isSavepoint()) { + return r1.props.isSavepoint() ? -1 : 1; + } else if (r1.props.forceCheckpoint() != r2.props.forceCheckpoint()) { + return r1.props.forceCheckpoint() ? -1 : 1; + } else if (r1.isPeriodic != r2.isPeriodic) { + return r1.isPeriodic ? 1 : -1; + } else { + return Long.compare(r1.timestamp, r2.timestamp); + } + }); + + CheckpointRequestDecider( + int maxConcurrentCheckpointAttempts, + Consumer<Long> rescheduleTrigger, + Clock clock, + long minPauseBetweenCheckpoints, + Supplier<Integer> pendingCheckpointsSizeSupplier, + Function<Supplier<Optional<CheckpointTriggerRequest>>, + Optional<CheckpointTriggerRequest>> sync) { + this.maxConcurrentCheckpointAttempts = maxConcurrentCheckpointAttempts; + this.rescheduleTrigger = rescheduleTrigger; + this.clock = clock; + this.minPauseBetweenCheckpoints = minPauseBetweenCheckpoints; + this.pendingCheckpointsSizeSupplier = pendingCheckpointsSizeSupplier; + this.sync = sync; + } + + Optional<CheckpointTriggerRequest> chooseRequestToExecute(CheckpointTriggerRequest newRequest, boolean isTriggering, long lastCompletionMs) { + return chooseRequestToExecute(false, () -> newRequest, isTriggering, lastCompletionMs); + } + + Optional<CheckpointTriggerRequest> chooseQueuedRequestToExecute(boolean isTriggering, long lastCompletionMs) { + return chooseRequestToExecute(true, triggerRequestQueue::peek, isTriggering, lastCompletionMs); + } + + /** + * Choose the next {@link CheckpointTriggerRequest request} to execute based on the provided candidate and the + * current state. Acquires a lock and may update the state. + * @return request to execute, if any. + */ + private Optional<CheckpointTriggerRequest> chooseRequestToExecute( + boolean isEnqueued, + Supplier<CheckpointTriggerRequest> candidateSupplier, Review comment: In that case I think it would be better to pull `synchronized (lock) {` to the upper methods (two package private ones), and leave the `private` internal method unsynchronized. ########## File path: flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/CheckpointRequestDecider.java ########## @@ -0,0 +1,156 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.flink.runtime.checkpoint; + +import org.apache.flink.annotation.VisibleForTesting; +import org.apache.flink.runtime.checkpoint.CheckpointCoordinator.CheckpointTriggerRequest; +import org.apache.flink.runtime.util.clock.Clock; + +import java.util.Optional; +import java.util.PriorityQueue; +import java.util.function.Consumer; +import java.util.function.Function; +import java.util.function.Supplier; + +import static org.apache.flink.runtime.checkpoint.CheckpointFailureReason.MINIMUM_TIME_BETWEEN_CHECKPOINTS; +import static org.apache.flink.runtime.checkpoint.CheckpointFailureReason.TOO_MANY_CONCURRENT_CHECKPOINTS; + +@SuppressWarnings("ConstantConditions") +class CheckpointRequestDecider { + private static final int MAX_QUEUED_REQUESTS = 1000; + + private final int maxConcurrentCheckpointAttempts; + private final Consumer<Long> rescheduleTrigger; + private final Clock clock; + private final long minPauseBetweenCheckpoints; + private final Supplier<Integer> pendingCheckpointsSizeSupplier; + private final Function<Supplier<Optional<CheckpointTriggerRequest>>, Optional<CheckpointTriggerRequest>> sync; + private final PriorityQueue<CheckpointTriggerRequest> triggerRequestQueue = new PriorityQueue<>((r1, r2) -> { + if (r1.props.isSavepoint() != r2.props.isSavepoint()) { + return r1.props.isSavepoint() ? -1 : 1; + } else if (r1.props.forceCheckpoint() != r2.props.forceCheckpoint()) { + return r1.props.forceCheckpoint() ? -1 : 1; + } else if (r1.isPeriodic != r2.isPeriodic) { + return r1.isPeriodic ? 1 : -1; + } else { + return Long.compare(r1.timestamp, r2.timestamp); + } + }); + + CheckpointRequestDecider( + int maxConcurrentCheckpointAttempts, + Consumer<Long> rescheduleTrigger, + Clock clock, + long minPauseBetweenCheckpoints, + Supplier<Integer> pendingCheckpointsSizeSupplier, + Function<Supplier<Optional<CheckpointTriggerRequest>>, + Optional<CheckpointTriggerRequest>> sync) { + this.maxConcurrentCheckpointAttempts = maxConcurrentCheckpointAttempts; + this.rescheduleTrigger = rescheduleTrigger; + this.clock = clock; + this.minPauseBetweenCheckpoints = minPauseBetweenCheckpoints; + this.pendingCheckpointsSizeSupplier = pendingCheckpointsSizeSupplier; + this.sync = sync; + } + + Optional<CheckpointTriggerRequest> chooseRequestToExecute(CheckpointTriggerRequest newRequest, boolean isTriggering, long lastCompletionMs) { + return chooseRequestToExecute(false, () -> newRequest, isTriggering, lastCompletionMs); + } + + Optional<CheckpointTriggerRequest> chooseQueuedRequestToExecute(boolean isTriggering, long lastCompletionMs) { + return chooseRequestToExecute(true, triggerRequestQueue::peek, isTriggering, lastCompletionMs); + } + + /** + * Choose the next {@link CheckpointTriggerRequest request} to execute based on the provided candidate and the + * current state. Acquires a lock and may update the state. + * @return request to execute, if any. + */ + private Optional<CheckpointTriggerRequest> chooseRequestToExecute( + boolean isEnqueued, + Supplier<CheckpointTriggerRequest> candidateSupplier, + boolean isTriggering, + long lastCompletionMs) { + return sync.apply(() -> { + final CheckpointTriggerRequest candidate = candidateSupplier.get(); + if (candidate == null) { + return Optional.empty(); Review comment: As above, moving lock to the callers would solve the problem. `candidate` (what's returned from the `Supplier`) is `@Nullable` implicitly (while none can tell it from the signature) ########## File path: flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/CheckpointCoordinator.java ########## @@ -535,20 +540,16 @@ private void startTriggeringCheckpoint(CheckpointTriggerRequest request) { request.getOnCompletionFuture()), timer); - final CompletableFuture<?> masterStatesComplete = pendingCheckpointCompletableFuture - .thenCompose(this::snapshotMasterState); + final CompletableFuture<?> masterStatesComplete = pendingCheckpointCompletableFuture.thenCompose(this::snapshotMasterState); - final CompletableFuture<?> coordinatorCheckpointsComplete = pendingCheckpointCompletableFuture - .thenComposeAsync((pendingCheckpoint) -> - OperatorCoordinatorCheckpoints.triggerAndAcknowledgeAllCoordinatorCheckpointsWithCompletion( - coordinatorsToCheckpoint, pendingCheckpoint, timer), - timer); + final CompletableFuture<?> coordinatorCheckpointsComplete = pendingCheckpointCompletableFuture.thenComposeAsync( + pendingCheckpoint -> triggerAndAcknowledgeAllCoordinatorCheckpointsWithCompletion(coordinatorsToCheckpoint, pendingCheckpoint, timer), + timer); Review comment: Hmmm, indeed it doesn't say anything. At least not for java (for Scala it says 100). Those lines are to wide for example for side-by-side comparisons (diffs) or when opening two files with vertical split. ########## File path: flink-streaming-java/src/main/java/org/apache/flink/streaming/api/graph/StreamingJobGraphGenerator.java ########## @@ -969,15 +969,17 @@ private void configureCheckpointing() { triggerVertices, ackVertices, commitVertices, - new CheckpointCoordinatorConfiguration( - interval, - cfg.getCheckpointTimeout(), - cfg.getMinPauseBetweenCheckpoints(), - cfg.getMaxConcurrentCheckpoints(), - retentionAfterTermination, - getCheckpointingMode(cfg) == CheckpointingMode.EXACTLY_ONCE, - cfg.isPreferCheckpointForRecovery(), - cfg.getTolerableCheckpointFailureNumber()), + CheckpointCoordinatorConfiguration.builder() + .setCheckpointInterval(interval) + .setCheckpointTimeout(cfg.getCheckpointTimeout()) + .setMinPauseBetweenCheckpoints(cfg.getMinPauseBetweenCheckpoints()) + .setMaxConcurrentCheckpoints(cfg.getMaxConcurrentCheckpoints()) + .setCheckpointRetentionPolicy(retentionAfterTermination) + .setExactlyOnce(getCheckpointingMode(cfg) == CheckpointingMode.EXACTLY_ONCE) + .setPreferCheckpointForRecovery(cfg.isPreferCheckpointForRecovery()) + .setTolerableCheckpointFailureNumber(cfg.getTolerableCheckpointFailureNumber()) + .setUnalignedCheckpointsEnabled(cfg.isUnalignedCheckpointsEnabled()) Review comment: Yes, you are right, never mind, I've misunderstood something. I think everything is fine here. ########## File path: flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/CheckpointRequestDecider.java ########## @@ -0,0 +1,156 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.flink.runtime.checkpoint; + +import org.apache.flink.annotation.VisibleForTesting; +import org.apache.flink.runtime.checkpoint.CheckpointCoordinator.CheckpointTriggerRequest; +import org.apache.flink.runtime.util.clock.Clock; + +import java.util.Optional; +import java.util.PriorityQueue; +import java.util.function.Consumer; +import java.util.function.Supplier; + +import static org.apache.flink.runtime.checkpoint.CheckpointFailureReason.MINIMUM_TIME_BETWEEN_CHECKPOINTS; +import static org.apache.flink.runtime.checkpoint.CheckpointFailureReason.TOO_MANY_CONCURRENT_CHECKPOINTS; + +@SuppressWarnings("ConstantConditions") +class CheckpointRequestDecider { + private static final int MAX_QUEUED_REQUESTS = 1000; + + private final int maxConcurrentCheckpointAttempts; + private final Consumer<Long> rescheduleTrigger; + private final Clock clock; + private final long minPauseBetweenCheckpoints; + private final Supplier<Integer> pendingCheckpointsSizeSupplier; + private final Object lock; + private final PriorityQueue<CheckpointTriggerRequest> triggerRequestQueue = new PriorityQueue<>((r1, r2) -> { + if (r1.props.isSavepoint() != r2.props.isSavepoint()) { + return r1.props.isSavepoint() ? -1 : 1; + } else if (r1.props.forceCheckpoint() != r2.props.forceCheckpoint()) { + return r1.props.forceCheckpoint() ? -1 : 1; + } else if (r1.isPeriodic != r2.isPeriodic) { + return r1.isPeriodic ? 1 : -1; + } else { + return Long.compare(r1.timestamp, r2.timestamp); + } + }); + + CheckpointRequestDecider( + int maxConcurrentCheckpointAttempts, + Consumer<Long> rescheduleTrigger, + Clock clock, + long minPauseBetweenCheckpoints, + Supplier<Integer> pendingCheckpointsSizeSupplier, + Object lock) { + this.maxConcurrentCheckpointAttempts = maxConcurrentCheckpointAttempts; + this.rescheduleTrigger = rescheduleTrigger; + this.clock = clock; + this.minPauseBetweenCheckpoints = minPauseBetweenCheckpoints; + this.pendingCheckpointsSizeSupplier = pendingCheckpointsSizeSupplier; + this.lock = lock; + } + + Optional<CheckpointTriggerRequest> chooseRequestToExecute(CheckpointTriggerRequest newRequest, boolean isTriggering, long lastCompletionMs) { + return chooseRequestToExecute(false, () -> newRequest, isTriggering, lastCompletionMs); + } + + Optional<CheckpointTriggerRequest> chooseQueuedRequestToExecute(boolean isTriggering, long lastCompletionMs) { + return chooseRequestToExecute(true, triggerRequestQueue::peek, isTriggering, lastCompletionMs); + } + + /** + * Choose the next {@link CheckpointTriggerRequest request} to execute based on the provided candidate and the + * current state. Acquires a lock and may update the state. + * @return request to execute, if any. + */ + private Optional<CheckpointTriggerRequest> chooseRequestToExecute( + boolean isEnqueued, + Supplier<CheckpointTriggerRequest> candidateSupplier, + boolean isTriggering, + long lastCompletionMs) { + synchronized (lock) { + final CheckpointTriggerRequest candidate = candidateSupplier.get(); + if (candidate == null) { + return Optional.empty(); + } else if (isTriggering) { + return onCantTriggerNow(candidate, isEnqueued); + } else if (pendingCheckpointsSizeSupplier.get() >= maxConcurrentCheckpointAttempts) { + return onCantCheckpointNow(candidate, TOO_MANY_CONCURRENT_CHECKPOINTS, () -> { /* not used */ }, isEnqueued); + } else if (nextTriggerDelayMillis(lastCompletionMs) > 0) { + return onCantCheckpointNow(candidate, MINIMUM_TIME_BETWEEN_CHECKPOINTS, () -> rescheduleTrigger.accept(nextTriggerDelayMillis(lastCompletionMs)), isEnqueued); + } else if (candidate != triggerRequestQueue.peek()) { + triggerRequestQueue.offer(candidate); + return Optional.of(triggerRequestQueue.poll()); + } else { + return Optional.of(isEnqueued ? triggerRequestQueue.poll() : candidate); + } + } + } + + private Optional<CheckpointTriggerRequest> onCantTriggerNow(CheckpointTriggerRequest candidate, boolean isEnqueued) { + offerIfNeeded(candidate, isEnqueued); + return Optional.empty(); + } + + private Optional<CheckpointTriggerRequest> onCantCheckpointNow( + CheckpointTriggerRequest candidate, + CheckpointFailureReason dropReason, + Runnable postDrop, + boolean isEnqueued) { + if (candidate.props.forceCheckpoint()) { + return Optional.of(isEnqueued ? triggerRequestQueue.poll() : candidate); Review comment: True, it would add an extra `PriorityQueue` access, but I would say it's a bit premature optimisation. For one thing it's not performance critical code - it's executed very rarely. Also costs of the lock acquisitions + other overheads would be far greater than the `PriorityQueue`. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
