accumulo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From joshelser <...@git.apache.org>
Subject [GitHub] accumulo pull request #224: ACCUMULO-4500 ACCUMULO-96 Added summarization
Date Thu, 09 Mar 2017 19:43:08 GMT
Github user joshelser commented on a diff in the pull request:

    https://github.com/apache/accumulo/pull/224#discussion_r105250934
  
    --- Diff: core/src/main/java/org/apache/accumulo/core/summary/Gatherer.java ---
    @@ -0,0 +1,513 @@
    +/*
    + * Licensed to the Apache Software Foundation (ASF) under one or more
    + * contributor license agreements.  See the NOTICE file distributed with
    + * this work for additional information regarding copyright ownership.
    + * The ASF licenses this file to You under the Apache License, Version 2.0
    + * (the "License"); you may not use this file except in compliance with
    + * the License.  You may obtain a copy of the License at
    + *
    + *     http://www.apache.org/licenses/LICENSE-2.0
    + *
    + * Unless required by applicable law or agreed to in writing, software
    + * distributed under the License is distributed on an "AS IS" BASIS,
    + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    + * See the License for the specific language governing permissions and
    + * limitations under the License.
    + */
    +
    +package org.apache.accumulo.core.summary;
    +
    +import java.io.IOException;
    +import java.io.UncheckedIOException;
    +import java.util.ArrayList;
    +import java.util.Collections;
    +import java.util.HashMap;
    +import java.util.HashSet;
    +import java.util.Iterator;
    +import java.util.List;
    +import java.util.Map;
    +import java.util.Map.Entry;
    +import java.util.Set;
    +import java.util.TreeMap;
    +import java.util.concurrent.Callable;
    +import java.util.concurrent.ExecutionException;
    +import java.util.concurrent.ExecutorService;
    +import java.util.concurrent.Executors;
    +import java.util.concurrent.Future;
    +import java.util.function.Predicate;
    +import java.util.regex.Pattern;
    +import java.util.stream.Collectors;
    +import java.util.stream.StreamSupport;
    +
    +import org.apache.accumulo.core.client.AccumuloException;
    +import org.apache.accumulo.core.client.AccumuloSecurityException;
    +import org.apache.accumulo.core.client.TableNotFoundException;
    +import org.apache.accumulo.core.client.impl.ClientContext;
    +import org.apache.accumulo.core.client.impl.ServerClient;
    +import org.apache.accumulo.core.client.impl.thrift.TRowRange;
    +import org.apache.accumulo.core.client.impl.thrift.TSummaries;
    +import org.apache.accumulo.core.client.impl.thrift.TSummaryRequest;
    +import org.apache.accumulo.core.client.summary.SummarizerConfiguration;
    +import org.apache.accumulo.core.conf.AccumuloConfiguration;
    +import org.apache.accumulo.core.data.ByteSequence;
    +import org.apache.accumulo.core.data.Key;
    +import org.apache.accumulo.core.data.Range;
    +import org.apache.accumulo.core.data.impl.KeyExtent;
    +import org.apache.accumulo.core.file.blockfile.cache.BlockCache;
    +import org.apache.accumulo.core.metadata.schema.MetadataScanner;
    +import org.apache.accumulo.core.metadata.schema.TabletMetadata;
    +import org.apache.accumulo.core.rpc.ThriftUtil;
    +import org.apache.accumulo.core.tabletserver.thrift.TabletClientService.Client;
    +import org.apache.accumulo.core.trace.Tracer;
    +import org.apache.accumulo.core.trace.thrift.TInfo;
    +import org.apache.accumulo.core.util.ByteBufferUtil;
    +import org.apache.accumulo.core.util.CachedConfiguration;
    +import org.apache.accumulo.core.util.TextUtil;
    +import org.apache.accumulo.fate.util.UtilWaitThread;
    +import org.apache.hadoop.conf.Configuration;
    +import org.apache.hadoop.fs.FileSystem;
    +import org.apache.hadoop.fs.Path;
    +import org.apache.hadoop.io.Text;
    +import org.apache.thrift.TApplicationException;
    +import org.apache.thrift.transport.TTransportException;
    +import org.slf4j.Logger;
    +import org.slf4j.LoggerFactory;
    +
    +import com.google.common.base.Preconditions;
    +import com.google.common.collect.Lists;
    +import com.google.common.hash.Hashing;
    +import com.google.common.net.HostAndPort;
    +
    +/**
    + * This class implements using multiple tservers to gather summaries.
    + *
    + * Below is a rough outline of the RPC process.
    + *
    + * <ol>
    + * <li>Clients pick a random tserver and make an RPC to remotely execute {@link
#gather()}.
    + * <li> {@link #gather()} will call make RPC calls to multiple tservers to remotely
execute {@link #processPartition(int, int)}
    + * <li> {@link #processPartition(int, int)} will make RPC calls to multiple tserver
to remotely execute
    + * {@link #processFiles(FileSystemResolver, Map, BlockCache, BlockCache, ExecutorService)}
    + * </ol>
    + */
    +public class Gatherer {
    +
    +  private static final Logger log = LoggerFactory.getLogger(Gatherer.class);
    +
    +  private ClientContext ctx;
    +  private String tableId;
    +  private SummarizerFactory factory;
    +  private Text startRow = null;
    +  private Text endRow = null;
    +  private Range clipRange;
    +  private Predicate<SummarizerConfiguration> summarySelector;
    +
    +  private TSummaryRequest request;
    +
    +  private String summarizerPattern;
    +
    +  private Set<SummarizerConfiguration> summaries;
    +
    +  public Gatherer(ClientContext context, TSummaryRequest request, AccumuloConfiguration
tableConfig) {
    +    this.ctx = context;
    +    this.tableId = request.tableId;
    +    this.startRow = ByteBufferUtil.toText(request.bounds.startRow);
    +    this.endRow = ByteBufferUtil.toText(request.bounds.endRow);
    +    this.clipRange = new Range(startRow, false, endRow, true);
    +    this.summaries = request.getSummarizers().stream().map(SummarizerConfigurationUtil::fromThrift).collect(Collectors.toSet());
    +    this.request = request;
    +
    +    this.summarizerPattern = request.getSummarizerPattern();
    +
    +    if (summarizerPattern != null) {
    +      Pattern pattern = Pattern.compile(summarizerPattern);
    +      // The way conf is converted to string below is documented in the API, so consider
this when making changes!
    +      summarySelector = conf -> pattern.matcher(conf.getClassName() + " " + new TreeMap<>(conf.getOptions())).matches();
    +      if (!summaries.isEmpty()) {
    +        summarySelector = summarySelector.or(conf -> summaries.contains(conf));
    +      }
    +    } else if (!summaries.isEmpty()) {
    +      summarySelector = conf -> summaries.contains(conf);
    +    } else {
    +      summarySelector = conf -> true;
    +    }
    +
    +    this.factory = new SummarizerFactory(tableConfig);
    +  }
    +
    +  private TSummaryRequest getRequest() {
    +    return request;
    +  }
    +
    +  /**
    +   * @param fileSelector
    +   *          only returns files that match this predicate
    +   * @return A map of the form : {@code map<tserver location, map<path, list<range>>}
. The ranges associated with a file represent the tablets that use the
    +   *         file.
    +   */
    +  private Map<String,Map<String,List<TRowRange>>> getFilesGroupedByLocation(Predicate<String>
fileSelector) throws TableNotFoundException, AccumuloException,
    +      AccumuloSecurityException {
    +
    +    Iterable<TabletMetadata> tmi = MetadataScanner.builder().from(ctx).overUserTableId(tableId,
startRow, endRow).fetchFiles().fetchLocation().fetchLast()
    +        .fetchPrev().build();
    +
    +    // get a subset of files
    +    Map<String,List<TabletMetadata>> files = new HashMap<>();
    +    for (TabletMetadata tm : tmi) {
    +      for (String file : tm.getFiles()) {
    +        if (fileSelector.test(file)) {
    +          // TODO push this filtering to server side and possibly use batch scanner
    +          files.computeIfAbsent(file, s -> new ArrayList<>()).add(tm);
    +        }
    +      }
    +    }
    +
    +    // group by location, then file
    +
    +    Map<String,Map<String,List<TRowRange>>> locations = new HashMap<>();
    +
    +    List<String> tservers = null;
    +
    +    for (Entry<String,List<TabletMetadata>> entry : files.entrySet()) {
    +
    +      String location = entry.getValue().stream().filter(tm -> tm.getLocation() !=
null) // filter tablets w/o a location
    +          .map(tm -> tm.getLocation().getHostAndPort().toString()) // convert to host:port
strings
    +          .min(String::compareTo) // find minimum host:port
    +          .orElse(entry.getValue().stream().filter(tm -> tm.getLast() != null) //
if no locations, then look at last locations
    +              .map(tm -> tm.getLast().getHostAndPort().toString()) // convert to host:port
strings
    +              .min(String::compareTo).orElse(null)); // find minimum last location or
return null
    +
    +      if (location == null) {
    +        if (tservers == null) {
    +          tservers = ctx.getConnector().instanceOperations().getTabletServers();
    +          Collections.sort(tservers);
    +        }
    +
    +        // When no location, the approach below will consistently choose the same tserver
for the same file (as long as the set of tservers is stable).
    +        int idx = Math.abs(Hashing.murmur3_32().hashString(entry.getKey()).asInt()) %
tservers.size();
    +        location = tservers.get(idx);
    +      }
    +
    +      List<Range> merged = Range.mergeOverlapping(Lists.transform(entry.getValue(),
tm -> tm.getExtent().toDataRange())); // merge contiguous ranges
    +      List<TRowRange> ranges = merged.stream().map(r -> toClippedExtent(r).toThrift()).collect(Collectors.toList());
// clip ranges to queried range
    +
    +      locations.computeIfAbsent(location, s -> new HashMap<>()).put(entry.getKey(),
ranges);
    +    }
    +
    +    return locations;
    +  }
    +
    +  private <K,V> Iterable<Map<K,V>> partition(Map<K,V> map, int
max) {
    +
    +    if (map.size() < max) {
    +      return Collections.singletonList(map);
    +    }
    +
    +    return new Iterable<Map<K,V>>() {
    +      @Override
    +      public Iterator<Map<K,V>> iterator() {
    +        Iterator<Entry<K,V>> esi = map.entrySet().iterator();
    +
    +        return new Iterator<Map<K,V>>() {
    +          @Override
    +          public boolean hasNext() {
    +            return esi.hasNext();
    +          }
    +
    +          @Override
    +          public Map<K,V> next() {
    +            Map<K,V> workingMap = new HashMap<>(max);
    +            while (esi.hasNext() && workingMap.size() < max) {
    +              Entry<K,V> entry = esi.next();
    +              workingMap.put(entry.getKey(), entry.getValue());
    +            }
    +            return workingMap;
    +          }
    +        };
    +      }
    +    };
    +  }
    +
    +  private static class ProcessedFiles {
    +    final SummaryCollection summaries;
    +    final Set<String> failedFiles;
    +
    +    public ProcessedFiles() {
    +      this.summaries = new SummaryCollection();
    +      this.failedFiles = new HashSet<>();
    +    }
    +  }
    +
    +  private class FilesProcessor implements Callable<ProcessedFiles> {
    +
    +    HostAndPort location;
    +    Map<String,List<TRowRange>> allFiles;
    +    private TInfo tinfo;
    +
    +    public FilesProcessor(TInfo tinfo, HostAndPort location, Map<String,List<TRowRange>>
allFiles) {
    +      this.location = location;
    +      this.allFiles = allFiles;
    +      this.tinfo = tinfo;
    +    }
    +
    +    @Override
    +    public ProcessedFiles call() throws Exception {
    +      ProcessedFiles pfiles = new ProcessedFiles();
    +
    +      Client client = null;
    +      try {
    +        client = ThriftUtil.getTServerClient(location, ctx);
    +        // partition files into smaller chunks so that not too many are sent to a tserver
at once
    +        for (Map<String,List<TRowRange>> files : partition(allFiles, 500))
{
    +          if (pfiles.failedFiles.size() > 0) {
    +            // there was a previous failure on this tserver, so just fail the rest of
the files
    +            pfiles.failedFiles.addAll(files.keySet());
    +            continue;
    +          }
    +
    +          TSummaries tSums = null;
    +          try {
    +            tSums = client.getSummariesFromFiles(tinfo, ctx.rpcCreds(), getRequest(),
files);
    +          } catch (TApplicationException tae) {
    +            throw tae;
    +          } catch (TTransportException e) {
    +            pfiles.failedFiles.addAll(files.keySet());
    +            continue;
    +          }
    +
    +          pfiles.summaries.merge(new SummaryCollection(tSums), factory);
    +        }
    +      } finally {
    +        ThriftUtil.returnClient(client);
    +      }
    +
    +      return pfiles;
    +    }
    +  }
    +
    +  /**
    +   * This methods reads a subset of file paths into memory and groups them by location.
Then it request sumaries for files from each location/tablet server.
    +   */
    +  public SummaryCollection processPartition(int modulus, int remainder) throws TableNotFoundException,
AccumuloException, AccumuloSecurityException {
    +
    +    Predicate<String> hashingFileSelector = file -> Math.abs(Hashing.murmur3_32().hashString(file).asInt())
% modulus == remainder;
    +
    +    Map<String,Map<String,List<TRowRange>>> filesGBL = getFilesGroupedByLocation(hashingFileSelector);
    +
    +    SummaryCollection partitionSummaries = new SummaryCollection();
    +
    +    ExecutorService execSrv = Executors.newCachedThreadPool();
    +    try {
    +      TInfo tinfo = Tracer.traceInfo();
    +      while (true) {
    +        Set<String> failedFiles = new HashSet<>();
    +
    +        List<Future<ProcessedFiles>> futures = new ArrayList<>();
    +        for (Entry<String,Map<String,List<TRowRange>>> entry : filesGBL.entrySet())
{
    +          HostAndPort location = HostAndPort.fromString(entry.getKey());
    +          Map<String,List<TRowRange>> allFiles = entry.getValue();
    +
    +          Future<ProcessedFiles> future = execSrv.submit(new FilesProcessor(tinfo,
location, allFiles));
    --- End diff --
    
    From `Executors.newCachedThreadPool()` javadoc:
    
    > These pools will typically improve the performance of programs that execute many
short-lived asynchronous tasks
    
    But, in `FilesProcessor` aren't we running an RPC without timeout? Seem to be competing
goals (and potentially keep us to hold on to a lot of threads if RPCs are delayed).


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

Mime
View raw message