1
0

[HUDI-3899] Drop index to delete pending index instants from timeline if applicable (#5342)

Co-authored-by: sivabalan <n.siva.b@gmail.com>
This commit is contained in:
Sagar Sumit
2022-04-19 07:58:46 +05:30
committed by GitHub
parent 52d878c52b
commit 4f44e6aeb5
4 changed files with 225 additions and 17 deletions

View File

@@ -18,13 +18,9 @@
package org.apache.hudi.metadata;
import org.apache.avro.specific.SpecificRecordBase;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileStatus;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hudi.avro.model.HoodieCleanMetadata;
import org.apache.hudi.avro.model.HoodieIndexPartitionInfo;
import org.apache.hudi.avro.model.HoodieIndexPlan;
import org.apache.hudi.avro.model.HoodieInstantInfo;
import org.apache.hudi.avro.model.HoodieMetadataRecord;
import org.apache.hudi.avro.model.HoodieRestoreMetadata;
@@ -70,6 +66,12 @@ import org.apache.hudi.config.metrics.HoodieMetricsJmxConfig;
import org.apache.hudi.exception.HoodieException;
import org.apache.hudi.exception.HoodieIndexException;
import org.apache.hudi.exception.HoodieMetadataException;
import org.apache.avro.specific.SpecificRecordBase;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileStatus;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.log4j.LogManager;
import org.apache.log4j.Logger;
@@ -88,6 +90,9 @@ import java.util.Set;
import java.util.stream.Collectors;
import static org.apache.hudi.common.table.HoodieTableConfig.ARCHIVELOG_FOLDER;
import static org.apache.hudi.common.table.timeline.HoodieInstant.State.REQUESTED;
import static org.apache.hudi.common.table.timeline.HoodieTimeline.getIndexInflightInstant;
import static org.apache.hudi.common.table.timeline.TimelineMetadataUtils.deserializeIndexPlan;
import static org.apache.hudi.common.util.StringUtils.EMPTY_STRING;
import static org.apache.hudi.metadata.HoodieTableMetadata.METADATA_TABLE_NAME_SUFFIX;
import static org.apache.hudi.metadata.HoodieTableMetadata.SOLO_COMMIT_TIMESTAMP;
@@ -727,9 +732,33 @@ public abstract class HoodieBackedTableMetadataWriter implements HoodieTableMeta
HoodieTableConfig.update(dataMetaClient.getFs(), new Path(dataMetaClient.getMetaPath()), dataMetaClient.getTableConfig().getProps());
LOG.warn("Deleting Metadata Table partitions: " + partitionPath);
dataMetaClient.getFs().delete(new Path(metadataWriteConfig.getBasePath(), partitionPath), true);
// delete corresponding pending indexing instant file in the timeline
LOG.warn("Deleting pending indexing instant from the timeline for partition: " + partitionPath);
deletePendingIndexingInstant(dataMetaClient, partitionPath);
}
}
/**
* Deletes any pending indexing instant, if it exists.
* It reads the plan from indexing.requested file and deletes both requested and inflight instants,
* if the partition path in the plan matches with the given partition path.
*/
private static void deletePendingIndexingInstant(HoodieTableMetaClient metaClient, String partitionPath) {
metaClient.reloadActiveTimeline().filterPendingIndexTimeline().getInstants().filter(instant -> REQUESTED.equals(instant.getState()))
.forEach(instant -> {
try {
HoodieIndexPlan indexPlan = deserializeIndexPlan(metaClient.getActiveTimeline().readIndexPlanAsBytes(instant).get());
if (indexPlan.getIndexPartitionInfos().stream()
.anyMatch(indexPartitionInfo -> indexPartitionInfo.getMetadataPartitionPath().equals(partitionPath))) {
metaClient.getActiveTimeline().deleteInstantFileIfExists(instant);
metaClient.getActiveTimeline().deleteInstantFileIfExists(getIndexInflightInstant(instant.getTimestamp()));
}
} catch (IOException e) {
LOG.error("Failed to delete the instant file corresponding to " + instant);
}
});
}
private MetadataRecordsGenerationParams getRecordsGenerationParams() {
return new MetadataRecordsGenerationParams(
dataMetaClient,