[HUDI-3365] Make sure Metadata Table records are updated appropriately on HDFS (#4739)

- This change makes sure MT records are updated appropriately on HDFS: previously after Log File append operations MT records were updated w/ just the size of the deltas being appended to the original files, which have been found to be the cause of issues in case of Rollbacks that were instead updating MT with records bearing the full file-size. - To make sure that we hedge against similar issues going f/w, this PR alleviates this discrepancy and streamlines the flow of MT table always ingesting records bearing full file-sizes.
2022-03-07 12:38:27 -08:00
parent f0bcee3c01
commit a66fd40692
18 changed files with 415 additions and 255 deletions
--- a/hudi-client/hudi-client-common/src/main/java/org/apache/hudi/client/BaseHoodieWriteClient.java
+++ b/hudi-client/hudi-client-common/src/main/java/org/apache/hudi/client/BaseHoodieWriteClient.java
@@ -18,6 +18,8 @@

 package org.apache.hudi.client;

+import com.codahale.metrics.Timer;
+import org.apache.hadoop.conf.Configuration;
 import org.apache.hudi.async.AsyncArchiveService;
 import org.apache.hudi.async.AsyncCleanerService;
 import org.apache.hudi.avro.model.HoodieCleanMetadata;
@@ -45,6 +47,7 @@ import org.apache.hudi.common.model.HoodieWriteStat;
 import org.apache.hudi.common.model.TableServiceType;
 import org.apache.hudi.common.model.WriteOperationType;
 import org.apache.hudi.common.table.HoodieTableMetaClient;
+import org.apache.hudi.common.table.HoodieTableVersion;
 import org.apache.hudi.common.table.timeline.HoodieActiveTimeline;
 import org.apache.hudi.common.table.timeline.HoodieInstant;
 import org.apache.hudi.common.table.timeline.HoodieInstant.State;
@@ -73,9 +76,8 @@ import org.apache.hudi.table.action.HoodieWriteMetadata;
 import org.apache.hudi.table.action.rollback.RollbackUtils;
 import org.apache.hudi.table.action.savepoint.SavepointHelpers;
 import org.apache.hudi.table.marker.WriteMarkersFactory;
-
-import com.codahale.metrics.Timer;
-import org.apache.hadoop.conf.Configuration;
+import org.apache.hudi.table.upgrade.SupportsUpgradeDowngrade;
+import org.apache.hudi.table.upgrade.UpgradeDowngrade;
 import org.apache.log4j.LogManager;
 import org.apache.log4j.Logger;

@@ -107,15 +109,16 @@ public abstract class BaseHoodieWriteClient<T extends HoodieRecordPayload, I, K,
  private static final long serialVersionUID = 1L;
  private static final Logger LOG = LogManager.getLogger(BaseHoodieWriteClient.class);

-  protected final transient HoodieMetrics metrics;
  private final transient HoodieIndex<?, ?> index;
+  private final SupportsUpgradeDowngrade upgradeDowngradeHelper;
+  private transient WriteOperationType operationType;
+  private transient HoodieWriteCommitCallback commitCallback;

+  protected final transient HoodieMetrics metrics;
  protected transient Timer.Context writeTimer = null;
  protected transient Timer.Context compactionTimer;
  protected transient Timer.Context clusteringTimer;

-  private transient WriteOperationType operationType;
-  private transient HoodieWriteCommitCallback commitCallback;
  protected transient AsyncCleanerService asyncCleanerService;
  protected transient AsyncArchiveService asyncArchiveService;
  protected final TransactionManager txnManager;
@@ -125,25 +128,32 @@ public abstract class BaseHoodieWriteClient<T extends HoodieRecordPayload, I, K,
   * Create a write client, with new hudi index.
   * @param context HoodieEngineContext
   * @param writeConfig instance of HoodieWriteConfig
+   * @param upgradeDowngradeHelper engine-specific instance of {@link SupportsUpgradeDowngrade}
   */
  @Deprecated
-  public BaseHoodieWriteClient(HoodieEngineContext context, HoodieWriteConfig writeConfig) {
-    this(context, writeConfig, Option.empty());
+  public BaseHoodieWriteClient(HoodieEngineContext context,
+                               HoodieWriteConfig writeConfig,
+                               SupportsUpgradeDowngrade upgradeDowngradeHelper) {
+    this(context, writeConfig, Option.empty(), upgradeDowngradeHelper);
  }

  /**
   * Create a write client, allows to specify all parameters.
+   *
   * @param context         HoodieEngineContext
-   * @param writeConfig instance of HoodieWriteConfig
+   * @param writeConfig     instance of HoodieWriteConfig
   * @param timelineService Timeline Service that runs as part of write client.
   */
  @Deprecated
-  public BaseHoodieWriteClient(HoodieEngineContext context, HoodieWriteConfig writeConfig,
-                                   Option<EmbeddedTimelineService> timelineService) {
+  public BaseHoodieWriteClient(HoodieEngineContext context,
+                               HoodieWriteConfig writeConfig,
+                               Option<EmbeddedTimelineService> timelineService,
+                               SupportsUpgradeDowngrade upgradeDowngradeHelper) {
    super(context, writeConfig, timelineService);
    this.metrics = new HoodieMetrics(config);
    this.index = createIndex(writeConfig);
    this.txnManager = new TransactionManager(config, fs);
+    this.upgradeDowngradeHelper = upgradeDowngradeHelper;
  }

  protected abstract HoodieIndex<?, ?> createIndex(HoodieWriteConfig writeConfig);
@@ -291,7 +301,7 @@ public abstract class BaseHoodieWriteClient<T extends HoodieRecordPayload, I, K,
    if (config.getWriteConcurrencyMode().supportsOptimisticConcurrencyControl()) {
      throw new HoodieException("Cannot bootstrap the table in multi-writer mode");
    }
-    HoodieTable<T, I, K, O> table = getTableAndInitCtx(WriteOperationType.UPSERT, HoodieTimeline.METADATA_BOOTSTRAP_INSTANT_TS);
+    HoodieTable<T, I, K, O> table = initTable(WriteOperationType.UPSERT, Option.ofNullable(HoodieTimeline.METADATA_BOOTSTRAP_INSTANT_TS));
    rollbackFailedBootstrap();
    table.bootstrap(context, extraMetadata);
  }
@@ -299,7 +309,7 @@ public abstract class BaseHoodieWriteClient<T extends HoodieRecordPayload, I, K,
  /**
   * Main API to rollback failed bootstrap.
   */
-  public void rollbackFailedBootstrap() {
+  protected void rollbackFailedBootstrap() {
    LOG.info("Rolling back pending bootstrap if present");
    HoodieTable<T, I, K, O> table = createTable(config, hadoopConf, config.isMetadataTableEnabled());
    HoodieTimeline inflightTimeline = table.getMetaClient().getCommitsTimeline().filterPendingExcludingCompaction();
@@ -628,7 +638,7 @@ public abstract class BaseHoodieWriteClient<T extends HoodieRecordPayload, I, K,
   * @return true if the savepoint was restored to successfully
   */
  public void restoreToSavepoint(String savepointTime) {
-    HoodieTable<T, I, K, O> table = createTable(config, hadoopConf, config.isMetadataTableEnabled());
+    HoodieTable<T, I, K, O> table = initTable(WriteOperationType.UNKNOWN, Option.empty());
    SavepointHelpers.validateSavepointPresence(table, savepointTime);
    restoreToInstant(savepointTime);
    SavepointHelpers.validateSavepointRestore(table, savepointTime);
@@ -636,25 +646,11 @@ public abstract class BaseHoodieWriteClient<T extends HoodieRecordPayload, I, K,

  @Deprecated
  public boolean rollback(final String commitInstantTime) throws HoodieRollbackException {
-    HoodieTable<T, I, K, O> table = createTable(config, hadoopConf);
+    HoodieTable<T, I, K, O> table = initTable(WriteOperationType.UNKNOWN, Option.empty());
    Option<HoodiePendingRollbackInfo> pendingRollbackInfo = getPendingRollbackInfo(table.getMetaClient(), commitInstantTime);
    return rollback(commitInstantTime, pendingRollbackInfo, false);
  }

-  /**
-   * @Deprecated
-   * Rollback the inflight record changes with the given commit time. This
-   * will be removed in future in favor of {@link BaseHoodieWriteClient#restoreToInstant(String)}
-   * Adding this api for backwards compatability.
-   * @param commitInstantTime Instant time of the commit
-   * @param skipLocking if this is triggered by another parent transaction, locking can be skipped.
-   * @throws HoodieRollbackException if rollback cannot be performed successfully
-   */
-  @Deprecated
-  public boolean rollback(final String commitInstantTime, boolean skipLocking) throws HoodieRollbackException {
-    return rollback(commitInstantTime, Option.empty(), skipLocking);
-  }
-
  /**
   * @Deprecated
   * Rollback the inflight record changes with the given commit time. This
@@ -711,7 +707,7 @@ public abstract class BaseHoodieWriteClient<T extends HoodieRecordPayload, I, K,
    final String restoreInstantTime = HoodieActiveTimeline.createNewInstantTime();
    Timer.Context timerContext = metrics.getRollbackCtx();
    try {
-      HoodieTable<T, I, K, O> table = createTable(config, hadoopConf, config.isMetadataTableEnabled());
+      HoodieTable<T, I, K, O> table = initTable(WriteOperationType.UNKNOWN, Option.empty());
      Option<HoodieRestorePlan> restorePlanOption = table.scheduleRestore(context, restoreInstantTime, instantTime);
      if (restorePlanOption.isPresent()) {
        HoodieRestoreMetadata restoreMetadata = table.restore(context, restoreInstantTime, instantTime);
@@ -988,7 +984,7 @@ public abstract class BaseHoodieWriteClient<T extends HoodieRecordPayload, I, K,
  /**
   * Rollback all failed writes.
   */
-  public Boolean rollbackFailedWrites() {
+  protected Boolean rollbackFailedWrites() {
    return rollbackFailedWrites(false);
  }

@@ -996,7 +992,7 @@ public abstract class BaseHoodieWriteClient<T extends HoodieRecordPayload, I, K,
   * Rollback all failed writes.
   * @param skipLocking if this is triggered by another parent transaction, locking can be skipped.
   */
-  public Boolean rollbackFailedWrites(boolean skipLocking) {
+  protected Boolean rollbackFailedWrites(boolean skipLocking) {
    HoodieTable<T, I, K, O> table = createTable(config, hadoopConf);
    List<String> instantsToRollback = getInstantsToRollback(table.getMetaClient(), config.getFailedWritesCleanPolicy(), Option.empty());
    Map<String, Option<HoodiePendingRollbackInfo>> pendingRollbacks = getPendingRollbackInfos(table.getMetaClient());
@@ -1246,17 +1242,79 @@ public abstract class BaseHoodieWriteClient<T extends HoodieRecordPayload, I, K,
  }

  /**
-   * Get HoodieTable and init {@link Timer.Context}.
+   * Instantiates engine-specific instance of {@link HoodieTable} as well as performs necessary
+   * bootstrapping operations (for ex, validating whether Metadata Table has to be bootstrapped)
   *
-   * @param operationType write operation type
+   * NOTE: THIS OPERATION IS EXECUTED UNDER LOCK, THEREFORE SHOULD AVOID ANY OPERATIONS
+   *       NOT REQUIRING EXTERNAL SYNCHRONIZATION
+   *
+   * @param metaClient instance of {@link HoodieTableMetaClient}
   * @param instantTime current inflight instant time
-   * @return HoodieTable
+   * @return instantiated {@link HoodieTable}
   */
-  protected abstract HoodieTable<T, I, K, O> getTableAndInitCtx(WriteOperationType operationType, String instantTime);
+  protected abstract HoodieTable<T, I, K, O> doInitTable(HoodieTableMetaClient metaClient, Option<String> instantTime);

  /**
-   * Sets write schema from last instant since deletes may not have schema set in the config.
+   * Instantiates and initializes instance of {@link HoodieTable}, performing crucial bootstrapping
+   * operations such as:
+   *
+   * NOTE: This method is engine-agnostic and SHOULD NOT be overloaded, please check on
+   * {@link #doInitTable(HoodieTableMetaClient, Option<String>)} instead
+   *
+   * <ul>
+   *   <li>Checking whether upgrade/downgrade is required</li>
+   *   <li>Bootstrapping Metadata Table (if required)</li>
+   *   <li>Initializing metrics contexts</li>
+   * </ul>
   */
+  protected final HoodieTable<T, I, K, O> initTable(WriteOperationType operationType, Option<String> instantTime) {
+    HoodieTableMetaClient metaClient = createMetaClient(true);
+    // Setup write schemas for deletes
+    if (operationType == WriteOperationType.DELETE) {
+      setWriteSchemaForDeletes(metaClient);
+    }
+
+    HoodieTable<T, I, K, O> table;
+
+    this.txnManager.beginTransaction();
+    try {
+      tryUpgrade(metaClient, instantTime);
+      table = doInitTable(metaClient, instantTime);
+    } finally {
+      this.txnManager.endTransaction();
+    }
+
+    // Validate table properties
+    metaClient.validateTableProperties(config.getProps(), operationType);
+    // Make sure that FS View is in sync
+    table.getHoodieView().sync();
+
+    switch (operationType) {
+      case INSERT:
+      case INSERT_PREPPED:
+      case UPSERT:
+      case UPSERT_PREPPED:
+      case BULK_INSERT:
+      case BULK_INSERT_PREPPED:
+      case INSERT_OVERWRITE:
+      case INSERT_OVERWRITE_TABLE:
+        setWriteTimer(table);
+        break;
+      case CLUSTER:
+        clusteringTimer = metrics.getClusteringCtx();
+        break;
+      case COMPACT:
+        compactionTimer = metrics.getCompactionCtx();
+        break;
+      default:
+    }
+
+    return table;
+  }
+
+  /**
+     * Sets write schema from last instant since deletes may not have schema set in the config.
+     */
  protected void setWriteSchemaForDeletes(HoodieTableMetaClient metaClient) {
    try {
      HoodieActiveTimeline activeTimeline = metaClient.getActiveTimeline();
@@ -1301,4 +1359,33 @@ public abstract class BaseHoodieWriteClient<T extends HoodieRecordPayload, I, K,
    this.heartbeatClient.stop();
    this.txnManager.close();
  }
+
+  private void setWriteTimer(HoodieTable<T, I, K, O> table) {
+    String commitType = table.getMetaClient().getCommitActionType();
+    if (commitType.equals(HoodieTimeline.COMMIT_ACTION)) {
+      writeTimer = metrics.getCommitCtx();
+    } else if (commitType.equals(HoodieTimeline.DELTA_COMMIT_ACTION)) {
+      writeTimer = metrics.getDeltaCommitCtx();
+    }
+  }
+
+  private void tryUpgrade(HoodieTableMetaClient metaClient, Option<String> instantTime) {
+    UpgradeDowngrade upgradeDowngrade =
+        new UpgradeDowngrade(metaClient, config, context, upgradeDowngradeHelper);
+
+    if (upgradeDowngrade.needsUpgradeOrDowngrade(HoodieTableVersion.current())) {
+      // Ensure no inflight commits by setting EAGER policy and explicitly cleaning all failed commits
+      List<String> instantsToRollback = getInstantsToRollback(metaClient, HoodieFailedWritesCleaningPolicy.EAGER, instantTime);
+
+      Map<String, Option<HoodiePendingRollbackInfo>> pendingRollbacks = getPendingRollbackInfos(metaClient);
+      instantsToRollback.forEach(entry -> pendingRollbacks.putIfAbsent(entry, Option.empty()));
+
+      rollbackFailedWrites(pendingRollbacks, true);
+
+      new UpgradeDowngrade(metaClient, config, context, upgradeDowngradeHelper)
+          .run(HoodieTableVersion.current(), instantTime.orElse(null));
+
+      metaClient.reloadActiveTimeline();
+    }
+  }
 }
--- a/hudi-client/hudi-client-common/src/main/java/org/apache/hudi/io/HoodieAppendHandle.java
+++ b/hudi-client/hudi-client-common/src/main/java/org/apache/hudi/io/HoodieAppendHandle.java
@@ -417,6 +417,8 @@ public class HoodieAppendHandle<T extends HoodieRecordPayload, I, K, O> extends
        writer = null;

        // update final size, once for all log files
+        // TODO we can actually deduce file size purely from AppendResult (based on offset and size
+        //      of the appended block)
        for (WriteStatus status: statuses) {
          long logFileSize = FSUtils.getFileSize(fs, new Path(config.getBasePath(), status.getStat().getPath()));
          status.getStat().setFileSizeInBytes(logFileSize);
--- a/hudi-client/hudi-client-common/src/main/java/org/apache/hudi/io/HoodieCreateHandle.java
+++ b/hudi-client/hudi-client-common/src/main/java/org/apache/hudi/io/HoodieCreateHandle.java
@@ -18,6 +18,10 @@

 package org.apache.hudi.io;

+import org.apache.avro.Schema;
+import org.apache.avro.generic.GenericRecord;
+import org.apache.avro.generic.IndexedRecord;
+import org.apache.hadoop.fs.Path;
 import org.apache.hudi.client.WriteStatus;
 import org.apache.hudi.common.engine.TaskContextSupplier;
 import org.apache.hudi.common.fs.FSUtils;
@@ -35,11 +39,6 @@ import org.apache.hudi.exception.HoodieInsertException;
 import org.apache.hudi.io.storage.HoodieFileWriter;
 import org.apache.hudi.io.storage.HoodieFileWriterFactory;
 import org.apache.hudi.table.HoodieTable;
-
-import org.apache.avro.Schema;
-import org.apache.avro.generic.GenericRecord;
-import org.apache.avro.generic.IndexedRecord;
-import org.apache.hadoop.fs.Path;
 import org.apache.log4j.LogManager;
 import org.apache.log4j.Logger;

@@ -235,20 +234,14 @@ public class HoodieCreateHandle<T extends HoodieRecordPayload, I, K, O> extends
    stat.setPrevCommit(HoodieWriteStat.NULL_COMMIT);
    stat.setFileId(writeStatus.getFileId());
    stat.setPath(new Path(config.getBasePath()), path);
-    stat.setTotalWriteBytes(computeTotalWriteBytes());
-    stat.setFileSizeInBytes(computeFileSizeInBytes());
    stat.setTotalWriteErrors(writeStatus.getTotalErrorRecords());
+
+    long fileSize = FSUtils.getFileSize(fs, path);
+    stat.setTotalWriteBytes(fileSize);
+    stat.setFileSizeInBytes(fileSize);
+
    RuntimeStats runtimeStats = new RuntimeStats();
    runtimeStats.setTotalCreateTime(timer.endTimer());
    stat.setRuntimeStats(runtimeStats);
  }
-
-  protected long computeTotalWriteBytes() throws IOException {
-    return FSUtils.getFileSize(fs, path);
-  }
-
-  protected long computeFileSizeInBytes() throws IOException {
-    return FSUtils.getFileSize(fs, path);
-  }
-
 }