drill-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (DRILL-4846) Eliminate extra operations during metadata cache pruning
Date Tue, 16 Aug 2016 23:06:20 GMT

    [ https://issues.apache.org/jira/browse/DRILL-4846?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15423562#comment-15423562
] 

ASF GitHub Bot commented on DRILL-4846:
---------------------------------------

Github user jinfengni commented on a diff in the pull request:

    https://github.com/apache/drill/pull/569#discussion_r75037246
  
    --- Diff: exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/Metadata.java
---
    @@ -470,41 +483,38 @@ private ParquetTableMetadataBase readBlockMeta(String path) throws
IOException {
         mapper.configure(DeserializationFeature.FAIL_ON_UNKNOWN_PROPERTIES, false);
         FSDataInputStream is = fs.open(p);
     
    -    ParquetTableMetadataBase parquetTableMetadata = mapper.readValue(is, ParquetTableMetadataBase.class);
    -    logger.info("Took {} ms to read metadata from cache file", timer.elapsed(TimeUnit.MILLISECONDS));
    -    timer.stop();
    -    if (tableModified(parquetTableMetadata, p)) {
    -      parquetTableMetadata =
    -          (createMetaFilesRecursively(Path.getPathWithoutSchemeAndAuthority(p.getParent()).toString())).getLeft();
    -    }
    -    return parquetTableMetadata;
    -  }
    -
    -  private ParquetTableMetadataDirs readMetadataDirs(String path) throws IOException {
    -    Stopwatch timer = Stopwatch.createStarted();
    -    Path p = new Path(path);
    -    ObjectMapper mapper = new ObjectMapper();
    -
    -    final SimpleModule serialModule = new SimpleModule();
    -    serialModule.addDeserializer(SchemaPath.class, new SchemaPath.De());
    -
    -    AfterburnerModule module = new AfterburnerModule();
    -    module.setUseOptimizedBeanDeserializer(true);
    +    boolean alreadyCheckedModification = false;
    +    boolean newMetadata = false;
     
    -    mapper.registerModule(serialModule);
    -    mapper.registerModule(module);
    -    mapper.configure(DeserializationFeature.FAIL_ON_UNKNOWN_PROPERTIES, false);
    -    FSDataInputStream is = fs.open(p);
    +    if (metaContext != null) {
    +      alreadyCheckedModification = metaContext.getStatus(parentDir.toString());
    +    }
     
    -    ParquetTableMetadataDirs parquetTableMetadataDirs = mapper.readValue(is, ParquetTableMetadataDirs.class);
    -    logger.info("Took {} ms to read directories from directory cache file", timer.elapsed(TimeUnit.MILLISECONDS));
    -    timer.stop();
    +    if (dirsOnly) {
    +      parquetTableMetadataDirs = mapper.readValue(is, ParquetTableMetadataDirs.class);
    +      logger.info("Took {} ms to read directories from directory cache file", timer.elapsed(TimeUnit.MILLISECONDS));
    +      timer.stop();
    +      if (!alreadyCheckedModification && tableModified(parquetTableMetadataDirs.getDirectories(),
p, parentDir, metaContext)) {
    --- End diff --
    
    tableModified compares the directory's modification time against metadata cache file's
modification time. 
    
    In 1st call,  the metadata file is .drill.parquet_metadata_directories. And we keep track
of the directory compared. In 2nd call, supposedly we should compare .drill.parquet_metadata
against directory. But MetaContext keeps track the comparison based on .drill.parquet_metadata_directories.
 Will this cause any potential problem, since .drill.parquet_metadata and .drill.parquet_metadata_directories
are different files?
     
    



> Eliminate extra operations during metadata cache pruning
> --------------------------------------------------------
>
>                 Key: DRILL-4846
>                 URL: https://issues.apache.org/jira/browse/DRILL-4846
>             Project: Apache Drill
>          Issue Type: Bug
>          Components: Metadata
>    Affects Versions: 1.7.0
>            Reporter: Aman Sinha
>            Assignee: Aman Sinha
>             Fix For: 1.8.0
>
>
> While doing performance testing for DRILL-4530 using a new data set and queries, we found
two potential performance issues: (a) the metadata cache was being read twice in some cases
and (b) the checking for directory modification time was being done twice, once as part of
the first phase of directory-based pruning and subsequently after the second phase pruning.
  This check gets expensive for large number of directories.   Creating this JIRA to track
fixes for these issues. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message