[FLINK-11159] Allow configuration whether to fall back to savepoints …

…for restore

Closes apache#8410

docs/dev/stream/state/checkpointing.md

@@ -76,6 +76,8 @@ Other parameters for checkpointing include:
 
  - *fail/continue task on checkpoint errors*: This determines if a task will be failed if an error occurs in the execution of the task's checkpoint procedure. This is the default behaviour. Alternatively, when this is disabled, the task will simply decline the checkpoint to the checkpoint coordinator and continue running.
 
+ - *prefer checkpoint for recovery*: This determines if a job will fallback to latest checkpoint even when there are more recent savepoints available to potentially reduce recovery time.
+
 <div class="codetabs" markdown="1">
 <div data-lang="java" markdown="1">
 {% highlight java %}
@@ -100,6 +102,9 @@ env.getCheckpointConfig().setMaxConcurrentCheckpoints(1);
 
 // enable externalized checkpoints which are retained after job cancellation
 env.getCheckpointConfig().enableExternalizedCheckpoints(ExternalizedCheckpointCleanup.RETAIN_ON_CANCELLATION);
+
+// allow job recovery fallback to checkpoint when there is a more recent savepoint
+env.getCheckpointConfig().setPreferCheckpointForRecovery(true);
 {% endhighlight %}
 </div>
 <div data-lang="scala" markdown="1">

docs/dev/stream/state/checkpointing.zh.md

@@ -76,6 +76,8 @@ Other parameters for checkpointing include:
 
  - *fail/continue task on checkpoint errors*: This determines if a task will be failed if an error occurs in the execution of the task's checkpoint procedure. This is the default behaviour. Alternatively, when this is disabled, the task will simply decline the checkpoint to the checkpoint coordinator and continue running.
 
+ - *prefer checkpoint for recovery*: This determines if a job will fallback to latest checkpoint even when there are more recent savepoints available to potentially reduce recovery time.
+
 <div class="codetabs" markdown="1">
 <div data-lang="java" markdown="1">
 {% highlight java %}
@@ -125,6 +127,9 @@ env.getCheckpointConfig.setFailTasksOnCheckpointingErrors(false)
 
 // allow only one checkpoint to be in progress at the same time
 env.getCheckpointConfig.setMaxConcurrentCheckpoints(1)
+
+// allow job recovery fallback to checkpoint when there is a more recent savepoint
+env.getCheckpointConfig().setPreferCheckpointForRecovery(true)
 {% endhighlight %}
 </div>
 </div>

flink-metrics/flink-metrics-jmx/src/test/java/org/apache/flink/runtime/jobmanager/JMXJobManagerMetricTest.java

@@ -100,7 +100,8 @@ public void testJobManagerJMXMetricAccess() throws Exception {
  50,
  5,
  CheckpointRetentionPolicy.NEVER_RETAIN_AFTER_TERMINATION,
- true),
+ true,
+ false),
  null));
 
  ClusterClient<?> client = MINI_CLUSTER_RESOURCE.getClusterClient();

flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/CheckpointCoordinator.java

@@ -183,6 +183,8 @@ public class CheckpointCoordinator {
  /** Registry that tracks state which is shared across (incremental) checkpoints. */
  private SharedStateRegistry sharedStateRegistry;
 
+ private boolean isPreferCheckpointForRecovery;
+
  // --------------------------------------------------------------------------------------------
 
  public CheckpointCoordinator(
@@ -199,7 +201,8 @@ public CheckpointCoordinator(
  CompletedCheckpointStore completedCheckpointStore,
  StateBackend checkpointStateBackend,
  Executor executor,
- SharedStateRegistryFactory sharedStateRegistryFactory) {
+ SharedStateRegistryFactory sharedStateRegistryFactory,
+ boolean isPreferCheckpointForRecovery) {
 
  // sanity checks
  checkNotNull(checkpointStateBackend);
@@ -233,6 +236,7 @@ public CheckpointCoordinator(
  this.executor = checkNotNull(executor);
  this.sharedStateRegistryFactory = checkNotNull(sharedStateRegistryFactory);
  this.sharedStateRegistry = sharedStateRegistryFactory.create(executor);
+ this.isPreferCheckpointForRecovery = isPreferCheckpointForRecovery;
 
  this.recentPendingCheckpoints = new ArrayDeque<>(NUM_GHOST_CHECKPOINT_IDS);
  this.masterHooks = new HashMap<>();
@@ -1080,7 +1084,7 @@ public boolean restoreLatestCheckpointedState(
  LOG.debug("Status of the shared state registry of job {} after restore: {}.", job, sharedStateRegistry);
 
  // Restore from the latest checkpoint
- CompletedCheckpoint latest = completedCheckpointStore.getLatestCheckpoint();
+ CompletedCheckpoint latest = completedCheckpointStore.getLatestCheckpoint(isPreferCheckpointForRecovery);
 
  if (latest == null) {
  if (errorIfNoCheckpoint) {

flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/CompletedCheckpointStore.java

@@ -19,18 +19,25 @@
 package org.apache.flink.runtime.checkpoint;
 
 import org.apache.flink.runtime.jobgraph.JobStatus;
+import org.apache.flink.util.FlinkRuntimeException;
+
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
 
 import java.util.List;
+import java.util.ListIterator;
 
 /**
  * A bounded LIFO-queue of {@link CompletedCheckpoint} instances.
  */
 public interface CompletedCheckpointStore {
 
+ Logger LOG = LoggerFactory.getLogger(CompletedCheckpointStore.class);
+
  /**
  * Recover available {@link CompletedCheckpoint} instances.
  *
- * <p>After a call to this method, {@link #getLatestCheckpoint()} returns the latest
+ * <p>After a call to this method, {@link #getLatestCheckpoint(boolean)} returns the latest
  * available checkpoint.
  */
  void recover() throws Exception;
@@ -47,7 +54,33 @@ public interface CompletedCheckpointStore {
  * Returns the latest {@link CompletedCheckpoint} instance or <code>null</code> if none was
  * added.
  */
- CompletedCheckpoint getLatestCheckpoint() throws Exception;
+ default CompletedCheckpoint getLatestCheckpoint(boolean isPreferCheckpointForRecovery) throws Exception {
+ if (getAllCheckpoints().isEmpty()) {
+ return null;
+ }
+
+ CompletedCheckpoint candidate = getAllCheckpoints().get(getAllCheckpoints().size() - 1);
+ if (isPreferCheckpointForRecovery && getAllCheckpoints().size() > 1) {
+ List<CompletedCheckpoint> allCheckpoints;
+ try {
+ allCheckpoints = getAllCheckpoints();
+ ListIterator<CompletedCheckpoint> listIterator = allCheckpoints.listIterator(allCheckpoints.size() - 1);
+ while (listIterator.hasPrevious()) {
+ CompletedCheckpoint prev = listIterator.previous();
+ if (!prev.getProperties().isSavepoint()) {
+ candidate = prev;
+ LOG.info("Found a completed checkpoint before the latest savepoint, will use it to recover!");
+ break;
+ }
+ }
+ } catch (Exception e) {
+ LOG.error("Method getAllCheckpoints caused exception : ", e);
+ throw new FlinkRuntimeException(e);
+ }
+ }
+
+ return candidate;
+ }
 
  /**
  * Shuts down the store.

flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/StandaloneCompletedCheckpointStore.java

@@ -76,11 +76,6 @@ public void addCheckpoint(CompletedCheckpoint checkpoint) throws Exception {
  }
  }
 
- @Override
- public CompletedCheckpoint getLatestCheckpoint() {
- return checkpoints.isEmpty() ? null : checkpoints.getLast();
- }
-
  @Override
  public List<CompletedCheckpoint> getAllCheckpoints() {
  return new ArrayList<>(checkpoints);

flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/ZooKeeperCompletedCheckpointStore.java

@@ -243,16 +243,6 @@ private void tryRemoveCompletedCheckpoint(CompletedCheckpoint completedCheckpoin
  }
  }
 
- @Override
- public CompletedCheckpoint getLatestCheckpoint() {
- if (completedCheckpoints.isEmpty()) {
- return null;
- }
- else {
- return completedCheckpoints.peekLast();
- }
- }
-
  @Override
  public List<CompletedCheckpoint> getAllCheckpoints() throws Exception {
  List<CompletedCheckpoint> checkpoints = new ArrayList<>(completedCheckpoints);

flink-runtime/src/main/java/org/apache/flink/runtime/executiongraph/ExecutionGraph.java

@@ -524,7 +524,8 @@ public void enableCheckpointing(
  CheckpointIDCounter checkpointIDCounter,
  CompletedCheckpointStore checkpointStore,
  StateBackend checkpointStateBackend,
- CheckpointStatsTracker statsTracker) {
+ CheckpointStatsTracker statsTracker,
+ boolean isPreferCheckpointForRecovery) {
 
  // simple sanity checks
  checkArgument(interval >= 10, "checkpoint interval must not be below 10ms");
@@ -554,7 +555,8 @@ public void enableCheckpointing(
  checkpointStore,
  checkpointStateBackend,
  ioExecutor,
- SharedStateRegistry.DEFAULT_FACTORY);
+ SharedStateRegistry.DEFAULT_FACTORY,
+ isPreferCheckpointForRecovery);
 
  // register the master hooks on the checkpoint coordinator
  for (MasterTriggerRestoreHook<?> hook : masterHooks) {

flink-runtime/src/main/java/org/apache/flink/runtime/executiongraph/ExecutionGraphBuilder.java

@@ -301,7 +301,8 @@ public static ExecutionGraph buildGraph(
  checkpointIdCounter,
  completedCheckpoints,
  rootBackend,
- checkpointStatsTracker);
+ checkpointStatsTracker,
+ chkConfig.isPreferCheckpointForRecovery());
  }
 
  // create all the metrics for the Execution Graph

flink-runtime/src/main/java/org/apache/flink/runtime/jobgraph/tasks/CheckpointCoordinatorConfiguration.java

@@ -54,13 +54,16 @@ public class CheckpointCoordinatorConfiguration implements Serializable {
  */
  private final boolean isExactlyOnce;
 
+ private final boolean isPreferCheckpointForRecovery;
+
  public CheckpointCoordinatorConfiguration(
  long checkpointInterval,
  long checkpointTimeout,
  long minPauseBetweenCheckpoints,
  int maxConcurrentCheckpoints,
  CheckpointRetentionPolicy checkpointRetentionPolicy,
- boolean isExactlyOnce) {
+ boolean isExactlyOnce,
+ boolean isPerfetCheckpointForRecovery) {
 
  // sanity checks
  if (checkpointInterval < 1 || checkpointTimeout < 1 ||
@@ -74,6 +77,7 @@ public CheckpointCoordinatorConfiguration(
  this.maxConcurrentCheckpoints = maxConcurrentCheckpoints;
  this.checkpointRetentionPolicy = Preconditions.checkNotNull(checkpointRetentionPolicy);
  this.isExactlyOnce = isExactlyOnce;
+ this.isPreferCheckpointForRecovery = isPerfetCheckpointForRecovery;
  }
 
  public long getCheckpointInterval() {
@@ -100,6 +104,10 @@ public boolean isExactlyOnce() {
  return isExactlyOnce;
  }
 
+ public boolean isPreferCheckpointForRecovery() {
+ return isPreferCheckpointForRecovery;
+ }
+
  @Override
  public boolean equals(Object o) {
  if (this == o) {
@@ -114,7 +122,8 @@ public boolean equals(Object o) {
  minPauseBetweenCheckpoints == that.minPauseBetweenCheckpoints &&
  maxConcurrentCheckpoints == that.maxConcurrentCheckpoints &&
  isExactlyOnce == that.isExactlyOnce &&
- checkpointRetentionPolicy == that.checkpointRetentionPolicy;
+ checkpointRetentionPolicy == that.checkpointRetentionPolicy &&
+ isPreferCheckpointForRecovery == that.isPreferCheckpointForRecovery;
  }
 
  @Override
@@ -125,7 +134,8 @@ public int hashCode() {
  minPauseBetweenCheckpoints,
  maxConcurrentCheckpoints,
  checkpointRetentionPolicy,
- isExactlyOnce);
+ isExactlyOnce,
+ isPreferCheckpointForRecovery);
  }
 
  @Override

flink-runtime/src/test/java/org/apache/flink/runtime/checkpoint/CheckpointCoordinatorFailureTest.java

@@ -80,7 +80,8 @@ public void testFailingCompletedCheckpointStoreAdd() throws Exception {
  new FailingCompletedCheckpointStore(),
  new MemoryStateBackend(),
  Executors.directExecutor(),
- SharedStateRegistry.DEFAULT_FACTORY);
+ SharedStateRegistry.DEFAULT_FACTORY,
+ false);
 
  coord.triggerCheckpoint(triggerTimestamp, false);
 
@@ -142,7 +143,7 @@ public void addCheckpoint(CompletedCheckpoint checkpoint) throws Exception {
  }
 
  @Override
- public CompletedCheckpoint getLatestCheckpoint() throws Exception {
+ public CompletedCheckpoint getLatestCheckpoint(boolean isPreferCheckpointForRecovery) throws Exception {
  throw new UnsupportedOperationException("Not implemented.");
  }
 

flink-runtime/src/test/java/org/apache/flink/runtime/checkpoint/CheckpointCoordinatorMasterHooksTest.java

@@ -206,7 +206,7 @@ public void testHooksAreCalledOnTrigger() throws Exception {
  assertEquals(0, cc.getNumberOfPendingCheckpoints());
 
  assertEquals(1, cc.getNumberOfRetainedSuccessfulCheckpoints());
- final CompletedCheckpoint chk = cc.getCheckpointStore().getLatestCheckpoint();
+ final CompletedCheckpoint chk = cc.getCheckpointStore().getLatestCheckpoint(false);
 
  final Collection<MasterState> masterStates = chk.getMasterHookStates();
  assertEquals(2, masterStates.size());
@@ -435,7 +435,8 @@ private static CheckpointCoordinator instantiateCheckpointCoordinator(JobID jid,
  new StandaloneCompletedCheckpointStore(10),
  new MemoryStateBackend(),
  Executors.directExecutor(),
- SharedStateRegistry.DEFAULT_FACTORY);
+ SharedStateRegistry.DEFAULT_FACTORY,
+ false);
  }
 
  private static <T> T mockGeneric(Class<?> clazz) {