Skip to content

[HUDI-4952] Fixing reading from metadata table when there are no inflight commits #6836

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
alexeykudinkin merged 5 commits into from
Oct 12, 2022
Merged

[HUDI-4952] Fixing reading from metadata table when there are no inflight commits #6836

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
34427d0
Fixing reading from metadata table when there are no inflight commits
nsivabalan Sep 29, 2022
23d923e
Fixing reading from metadata if not fully built out
nsivabalan Oct 3, 2022
e246d65
addressing minor comments
nsivabalan Oct 6, 2022
d7fbaa4
fixing sql conf and options interplay
nsivabalan Oct 7, 2022
8f763c8
addressing minor refactoring
nsivabalan Oct 10, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
35 changes: 29 additions & 6 deletions ...park-client/src/test/java/org/apache/hudi/client/functional/TestHoodieBackedMetadata.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -344,8 +344,8 @@ public void testTurnOffMetadataTableAfterEnable() throws Exception {
HoodieCommitMetadata hoodieCommitMetadata = doWriteOperationWithMeta(testTable, instant1, INSERT);

// Simulate the complete data directory including ".hoodie_partition_metadata" file
File metaForP1 = new File(metaClient.getBasePath() + "/p1",".hoodie_partition_metadata");
File metaForP2 = new File(metaClient.getBasePath() + "/p2",".hoodie_partition_metadata");
File metaForP1 = new File(metaClient.getBasePath() + "/p1", ".hoodie_partition_metadata");
File metaForP2 = new File(metaClient.getBasePath() + "/p2", ".hoodie_partition_metadata");
metaForP1.createNewFile();
metaForP2.createNewFile();

Expand Down Expand Up @@ -1716,8 +1716,8 @@ public void testMultiWriterForDoubleLocking() throws Exception {

HoodieWriteConfig writeConfig = getWriteConfigBuilder(true, true, false)
.withCleanConfig(HoodieCleanConfig.newBuilder()
.withFailedWritesCleaningPolicy(HoodieFailedWritesCleaningPolicy.LAZY).withAutoClean(true).retainCommits(4)
.build())
.withFailedWritesCleaningPolicy(HoodieFailedWritesCleaningPolicy.LAZY).withAutoClean(true).retainCommits(4)
.build())
.withAutoCommit(false)
.withWriteConcurrencyMode(WriteConcurrencyMode.OPTIMISTIC_CONCURRENCY_CONTROL)
.withLockConfig(HoodieLockConfig.newBuilder().withLockProvider(InProcessLockProvider.class).build())
Expand Down Expand Up @@ -1818,6 +1818,29 @@ public void testReattemptOfFailedClusteringCommit() throws Exception {
validateMetadata(client);
}

@Test
public void testMetadataReadWithNoCompletedCommits() throws Exception {
init(HoodieTableType.COPY_ON_WRITE);
HoodieSparkEngineContext engineContext = new HoodieSparkEngineContext(jsc);

List<HoodieRecord> records;
List<WriteStatus> writeStatuses;
String[] commitTimestamps = {HoodieActiveTimeline.createNewInstantTime(), HoodieActiveTimeline.createNewInstantTime()};

try (SparkRDDWriteClient client = new SparkRDDWriteClient(engineContext, getWriteConfig(true, true))) {
records = dataGen.generateInserts(commitTimestamps[0], 5);
client.startCommitWithTime(commitTimestamps[0]);
writeStatuses = client.bulkInsert(jsc.parallelize(records, 1), commitTimestamps[0]).collect();
assertNoWriteErrors(writeStatuses);

// make all commits to inflight in metadata table. Still read should go through, just that it may not return any data.
FileCreateUtils.deleteDeltaCommit(basePath + "/.hoodie/metadata/", commitTimestamps[0]);
FileCreateUtils.deleteDeltaCommit(basePath + " /.hoodie/metadata/", HoodieTableMetadata.SOLO_COMMIT_TIMESTAMP);
assertEquals(getAllFiles(metadata(client)).stream().map(p -> p.getName()).map(n -> FSUtils.getCommitTime(n)).collect(Collectors.toSet()).size(), 0);
}
}


/**
* Ensure that the reader only reads completed instants.
*
Expand Down Expand Up @@ -2050,7 +2073,7 @@ public void testRollbackDuringUpgradeForDoubleLocking() throws IOException, Inte
properties.setProperty(LockConfiguration.LOCK_ACQUIRE_WAIT_TIMEOUT_MS_PROP_KEY, "3000");
HoodieWriteConfig writeConfig = getWriteConfigBuilder(false, true, false)
.withCleanConfig(HoodieCleanConfig.newBuilder()
.withFailedWritesCleaningPolicy(HoodieFailedWritesCleaningPolicy.LAZY).withAutoClean(false).build())
.withFailedWritesCleaningPolicy(HoodieFailedWritesCleaningPolicy.LAZY).withAutoClean(false).build())
.withWriteConcurrencyMode(WriteConcurrencyMode.OPTIMISTIC_CONCURRENCY_CONTROL)
.withLockConfig(HoodieLockConfig.newBuilder().withLockProvider(InProcessLockProvider.class).build())
.withProperties(properties)
Expand Down Expand Up @@ -2078,7 +2101,7 @@ public void testRollbackDuringUpgradeForDoubleLocking() throws IOException, Inte
// set hoodie.table.version to 2 in hoodie.properties file
changeTableVersion(HoodieTableVersion.TWO);
writeConfig = getWriteConfigBuilder(true, true, false).withRollbackUsingMarkers(false).withCleanConfig(HoodieCleanConfig.newBuilder()
.withFailedWritesCleaningPolicy(HoodieFailedWritesCleaningPolicy.LAZY).withAutoClean(false).build())
.withFailedWritesCleaningPolicy(HoodieFailedWritesCleaningPolicy.LAZY).withAutoClean(false).build())
.withWriteConcurrencyMode(WriteConcurrencyMode.OPTIMISTIC_CONCURRENCY_CONTROL)
.withLockConfig(HoodieLockConfig.newBuilder().withLockProvider(InProcessLockProvider.class).build())
.withProperties(properties)
Expand Down
8 changes: 6 additions & 2 deletions hudi-common/src/main/java/org/apache/hudi/metadata/HoodieTableMetadataUtil.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1052,8 +1052,12 @@ private static List<FileSlice> getPartitionFileSlices(HoodieTableMetaClient meta
HoodieTableFileSystemView fsView = fileSystemView.orElse(getFileSystemView(metaClient));
Stream<FileSlice> fileSliceStream;
if (mergeFileSlices) {
fileSliceStream = fsView.getLatestMergedFileSlicesBeforeOrOn(
partition, metaClient.getActiveTimeline().filterCompletedInstants().lastInstant().get().getTimestamp());
if (metaClient.getActiveTimeline().filterCompletedInstants().lastInstant().isPresent()) {
fileSliceStream = fsView.getLatestMergedFileSlicesBeforeOrOn(
partition, metaClient.getActiveTimeline().filterCompletedInstants().lastInstant().get().getTimestamp());
} else {
return Collections.EMPTY_LIST;
}
} else {
fileSliceStream = fsView.getLatestFileSlices(partition);
}
Expand Down
18 changes: 11 additions & 7 deletions hudi-spark-datasource/hudi-spark-common/src/main/scala/org/apache/hudi/HoodieFileIndex.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@ import org.apache.hudi.common.util.StringUtils
import org.apache.hudi.exception.HoodieException
import org.apache.hudi.keygen.constant.KeyGeneratorOptions
import org.apache.hudi.keygen.{TimestampBasedAvroKeyGenerator, TimestampBasedKeyGenerator}
import org.apache.hudi.metadata.HoodieMetadataPayload
import org.apache.hudi.metadata.{HoodieMetadataPayload, HoodieTableMetadataUtil}
import org.apache.spark.internal.Logging
import org.apache.spark.sql.catalyst.InternalRow
import org.apache.spark.sql.catalyst.expressions.{And, Expression, Literal}
Expand Down Expand Up @@ -74,7 +74,7 @@ case class HoodieFileIndex(spark: SparkSession,
spark = spark,
metaClient = metaClient,
schemaSpec = schemaSpec,
configProperties = getConfigProperties(spark, options),
configProperties = getConfigProperties(spark, options, metaClient),
queryPaths = HoodieFileIndex.getQueryPaths(options),
specifiedQueryInstant = options.get(DataSourceReadOptions.TIME_TRAVEL_AS_OF_INSTANT.key).map(HoodieSqlCommonUtils.formatQueryInstant),
fileStatusCache = fileStatusCache
Expand Down Expand Up @@ -293,16 +293,20 @@ object HoodieFileIndex extends Logging {
schema.fieldNames.filter { colName => refs.exists(r => resolver.apply(colName, r.name)) }
}

def getConfigProperties(spark: SparkSession, options: Map[String, String]) = {
private def isFilesPartitionAvailable(metaClient: HoodieTableMetaClient): Boolean = {
metaClient.getTableConfig.getMetadataPartitions.contains(HoodieTableMetadataUtil.PARTITION_NAME_FILES)
}

def getConfigProperties(spark: SparkSession, options: Map[String, String], metaClient: HoodieTableMetaClient) = {
val sqlConf: SQLConf = spark.sessionState.conf
val properties = new TypedProperties()

// To support metadata listing via Spark SQL we allow users to pass the config via SQL Conf in spark session. Users
// would be able to run SET hoodie.metadata.enable=true in the spark sql session to enable metadata listing.
properties.setProperty(HoodieMetadataConfig.ENABLE.key(),
sqlConf.getConfString(HoodieMetadataConfig.ENABLE.key(),
HoodieMetadataConfig.DEFAULT_METADATA_ENABLE_FOR_READERS.toString))
properties.putAll(options.filter(p => p._2 != null).asJava)
val isMetadataFilesPartitionAvailable = isFilesPartitionAvailable(metaClient) && sqlConf.getConfString(HoodieMetadataConfig.ENABLE.key(),
HoodieMetadataConfig.DEFAULT_METADATA_ENABLE_FOR_READERS.toString).toBoolean
properties.setProperty(HoodieMetadataConfig.ENABLE.key(), String.valueOf(isMetadataFilesPartitionAvailable))
properties.putAll(options.filter(p => p._2 != null && !p._1.equals(HoodieMetadataConfig.ENABLE.key())).asJava)
Copy link
Contributor Author

@nsivabalan nsivabalan Oct 3, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@alexeykudinkin : there was a bug in here before. detected while testing it.

Copy link
Contributor

@alexeykudinkin alexeykudinkin Oct 4, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@nsivabalan no need to do this filtering, just reverse the order of putAll and setProperty

properties
}

Expand Down
2 changes: 1 addition & 1 deletion ...-spark-datasource/hudi-spark-common/src/main/scala/org/apache/hudi/cdc/HoodieCDCRDD.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -98,7 +98,7 @@ class HoodieCDCRDD(
metaClient.getTableConfig.cdcSupplementalLoggingMode
)

private val props = HoodieFileIndex.getConfigProperties(spark, Map.empty)
private val props = HoodieFileIndex.getConfigProperties(spark, Map.empty, metaClient)

protected val payloadProps: Properties = Option(metaClient.getTableConfig.getPreCombineField)
.map { preCombineField =>
Expand Down