lokeshj1703 commented on code in PR #11921:
URL: https://github.com/apache/hudi/pull/11921#discussion_r1775705064
##########
hudi-utilities/src/main/java/org/apache/hudi/utilities/HoodieMetadataTableValidator.java:
##########
@@ -914,6 +950,97 @@ private void validateAllColumnStats(
validate(metadataBasedColStats, fsBasedColStats, partitionPath, "column
stats");
}
+ private void validatePartitionStats(
+ HoodieMetadataValidationContext metadataTableBasedContext,
+ Set<String> baseDataFilesForCleaning,
+ List<String> allPartitions) throws Exception {
+
+ HoodieSparkEngineContext engineContext = new HoodieSparkEngineContext(jsc);
+ HoodieData<HoodieMetadataColumnStats> partitionStatsUsingColStats =
getPartitionStatsUsingColStats(metadataTableBasedContext,
+ baseDataFilesForCleaning, allPartitions, engineContext);
+
+ TableSchemaResolver schemaResolver = new TableSchemaResolver(metaClient);
+ boolean enableMetadataTable = true;
+ HoodieMetadataConfig metadataConfig = HoodieMetadataConfig.newBuilder()
+ .enable(enableMetadataTable)
+ .withMetadataIndexBloomFilter(enableMetadataTable)
+ .withMetadataIndexColumnStats(enableMetadataTable)
+ .withEnableRecordIndex(enableMetadataTable)
+ .build();
+ PartitionStatsIndexSupport partitionStatsIndexSupport =
+ new
PartitionStatsIndexSupport(engineContext.getSqlContext().sparkSession(),
AvroConversionUtils.convertAvroSchemaToStructType(schemaResolver.getTableAvroSchema()),
metadataConfig, metaClient,
+ false);
+ HoodieData<HoodieMetadataColumnStats> partitionStats =
+
partitionStatsIndexSupport.loadColumnStatsIndexRecords(JavaConverters.asScalaBufferConverter(metadataTableBasedContext.allColumnNameList).asScala().toSeq(),
false);
+ JavaRDD<HoodieMetadataColumnStats> diffRDD =
HoodieJavaRDD.getJavaRDD(partitionStats).subtract(HoodieJavaRDD.getJavaRDD(partitionStatsUsingColStats));
+ if (!diffRDD.isEmpty()) {
+ List<HoodieMetadataColumnStats> diff = diffRDD.collect();
+ Set<String> partitionPaths =
diff.stream().map(HoodieMetadataColumnStats::getFileName).collect(Collectors.toSet());
+ StringBuilder statDiffMsg = new StringBuilder();
+ for (String partitionPath : partitionPaths) {
+ List<HoodieMetadataColumnStats> diffPartitionStatsColStats =
partitionStatsUsingColStats.filter(stat ->
stat.getFileName().equals(partitionPath)).collectAsList();
+ List<HoodieMetadataColumnStats> diffPartitionStats =
partitionStats.filter(stat ->
stat.getFileName().equals(partitionPath)).collectAsList();
+ statDiffMsg.append(String.format("Partition stats from MDT: %s from
colstats: %s", Arrays.toString(diffPartitionStats.toArray()),
Arrays.toString(diffPartitionStatsColStats.toArray())));
+ }
+ throw new HoodieValidationException(String.format("Partition stats
validation failed diff: %s", statDiffMsg));
+ }
+ }
+
+ private HoodieData<HoodieMetadataColumnStats>
getPartitionStatsUsingColStats(HoodieMetadataValidationContext
metadataTableBasedContext, Set<String> baseDataFilesForCleaning,
Review Comment:
We need to rely on colstats so that log files are also accounted for.
Created https://issues.apache.org/jira/browse/HUDI-8260 to fix the validation
logic in col stats.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
