impala Datanode-side support for getVolumeBlockLocations() must also be enabled in the client config

Datanode-side support for getVolumeBlockLocations() must also be enabled in the client configuration

HDFS 源码分析

/**
 * Get block location information about a list of {@link HdfsBlockLocation}.
 * Used by {@link DistributedFileSystem#getFileBlockStorageLocations(List)} to
 * get {@link BlockStorageLocation}s for blocks returned by
 * {@link DistributedFileSystem#getFileBlockLocations(org.apache.hadoop.fs.FileStatus, long, long)}
 * .
 * 
 * This is done by making a round of RPCs to the associated datanodes, asking
 * the volume of each block replica. The returned array of
 * {@link BlockStorageLocation} expose this information as a
 * {@link VolumeId}.
 * 
 * @param blockLocations
 *          target blocks on which to query volume location information
 * @return volumeBlockLocations original block array augmented with additional
 *         volume location information for each replica.
 */
public BlockStorageLocation[] getBlockStorageLocations(
    List<BlockLocation> blockLocations) throws IOException,
    UnsupportedOperationException, InvalidBlockTokenException {
  if (!getConf().getHdfsBlocksMetadataEnabled) {
    throw new UnsupportedOperationException("Datanode-side support for " +
        "getVolumeBlockLocations() must also be enabled in the client " +
        "configuration.");
  }
  // Downcast blockLocations and fetch out required LocatedBlock(s)
  List<LocatedBlock> blocks = new ArrayList<LocatedBlock>();
  for (BlockLocation loc : blockLocations) {
    if (!(loc instanceof HdfsBlockLocation)) {
      throw new ClassCastException("DFSClient#getVolumeBlockLocations " +
          "expected to be passed HdfsBlockLocations");
    }
    HdfsBlockLocation hdfsLoc = (HdfsBlockLocation) loc;
    blocks.add(hdfsLoc.getLocatedBlock());
  }
  
  // Re-group the LocatedBlocks to be grouped by datanodes, with the values
  // a list of the LocatedBlocks on the datanode.
  Map<DatanodeInfo, List<LocatedBlock>> datanodeBlocks = 
      new LinkedHashMap<DatanodeInfo, List<LocatedBlock>>();
  for (LocatedBlock b : blocks) {
    for (DatanodeInfo info : b.getLocations()) {
      if (!datanodeBlocks.containsKey(info)) {
        datanodeBlocks.put(info, new ArrayList<LocatedBlock>());
      }
      List<LocatedBlock> l = datanodeBlocks.get(info);
      l.add(b);
    }
  }
      
  // Make RPCs to the datanodes to get volume locations for its replicas
  TraceScope scope =
    Trace.startSpan("getBlockStorageLocations", traceSampler);
  Map<DatanodeInfo, HdfsBlocksMetadata> metadatas;
  try {
    metadatas = BlockStorageLocationUtil.
        queryDatanodesForHdfsBlocksMetadata(conf, datanodeBlocks,
            getConf().getFileBlockStorageLocationsNumThreads,
            getConf().getFileBlockStorageLocationsTimeoutMs,
            getConf().connectToDnViaHostname);
    if (LOG.isTraceEnabled()) {
      LOG.trace("metadata returned: "
          + Joiner.on("\n").withKeyValueSeparator("=").join(metadatas));
    }
  } finally {
    scope.close();
  }
  
  // Regroup the returned VolumeId metadata to again be grouped by
  // LocatedBlock rather than by datanode
  Map<LocatedBlock, List<VolumeId>> blockVolumeIds = BlockStorageLocationUtil
      .associateVolumeIdsWithBlocks(blocks, metadatas);
  
  // Combine original BlockLocations with new VolumeId information
  BlockStorageLocation[] volumeBlockLocations = BlockStorageLocationUtil
      .convertToVolumeBlockLocations(blocks, blockVolumeIds);

  return volumeBlockLocations;
}

修改hdfs-site.xml

we need add below config to /etc/hadoop/conf/hdfs-site.xml
<property>
    <name>dfs.datanode.hdfs-blocks-metadata.enabled</name> 
    <value>true</value>
</property>
<property> 
    <name>dfs.block.local-path-access.user</name> 
    <value>impala</value>
</property>
<property>
    <name>dfs.client.file-block-storage-locations.timeout.millis</name>
    <value>60000</value>
</property>

重启服务 ！！！

service impala-state-store restart
service impala-catalog restart
service impala-server restart

me

我就是忘了重启服务，只重启了hdfs，导致卡在这1个小时了。。。。
因为我知道是配置问题，但是检查好几遍我hdfs配置都没问题，就是忘了impala。。。。。。