1
0

[HUDI 1308] Harden RFC-15 Implementation based on production testing (#2441)

Addresses leaks, perf degradation observed during testing. These were regressions from the original rfc-15 PoC implementation.

* Pass a single instance of HoodieTableMetadata everywhere
* Fix tests and add config for enabling metrics
 - Removed special casing of assumeDatePartitioning inside FSUtils#getAllPartitionPaths()
 - Consequently, IOException is never thrown and many files had to be adjusted
- More diligent handling of open file handles in metadata table
 - Added config for controlling reuse of connections
 - Added config for turning off fallback to listing, so we can see tests fail
 - Changed all ipf listing code to cache/amortize the open/close for better performance
 - Timelineserver also reuses connections, for better performance
 - Without timelineserver, when metadata table is opened from executors, reuse is not allowed
 - HoodieMetadataConfig passed into HoodieTableMetadata#create as argument.
 -  Fix TestHoodieBackedTableMetadata#testSync
This commit is contained in:
vinoth chandar
2021-01-19 21:20:28 -08:00
committed by GitHub
parent e23967b9e9
commit 5ca0625b27
55 changed files with 767 additions and 570 deletions

View File

@@ -86,9 +86,8 @@ public class DFSHoodieDatasetInputReader extends DFSDeltaInputReader {
protected List<String> getPartitions(Option<Integer> partitionsLimit) throws IOException {
// Using FSUtils.getFS here instead of metaClient.getFS() since we dont want to count these listStatus
// calls in metrics as they are not part of normal HUDI operation.
FileSystem fs = FSUtils.getFs(metaClient.getBasePath(), metaClient.getHadoopConf());
HoodieSparkEngineContext engineContext = new HoodieSparkEngineContext(jsc);
List<String> partitionPaths = FSUtils.getAllPartitionPaths(engineContext, fs, metaClient.getBasePath(),
List<String> partitionPaths = FSUtils.getAllPartitionPaths(engineContext, metaClient.getBasePath(),
HoodieMetadataConfig.DEFAULT_METADATA_ENABLE_FOR_READERS, HoodieMetadataConfig.DEFAULT_METADATA_VALIDATE, false);
// Sort partition so we can pick last N partitions by default
Collections.sort(partitionPaths);