[HDFS-16538] EC decoding failed due to not enough valid inputs

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.4.0, 3.2.4, 3.3.5
• Component/s: erasure-coding
• Labels: pull-request-available
• Assignee: qinyuren Assign to me
• Reporter: qinyuren
• Created: 13/Apr/22 08:47
• Updated: 19/Apr/22 08:32
• Resolved: 19/Apr/22 04:41

Description

Currently, we found this error if the #StripeReader.readStripe() have more than one block read failed.

We use the EC policy ec(6+3) in our cluster.

Caused by: org.apache.hadoop.HadoopIllegalArgumentException: No enough valid inputs are provided, not recoverable
        at org.apache.hadoop.io.erasurecode.rawcoder.ByteBufferDecodingState.checkInputBuffers(ByteBufferDecodingState.java:119)
        at org.apache.hadoop.io.erasurecode.rawcoder.ByteBufferDecodingState.<init>(ByteBufferDecodingState.java:47)
        at org.apache.hadoop.io.erasurecode.rawcoder.RawErasureDecoder.decode(RawErasureDecoder.java:86)
        at org.apache.hadoop.io.erasurecode.rawcoder.RawErasureDecoder.decode(RawErasureDecoder.java:170)
        at org.apache.hadoop.hdfs.StripeReader.decodeAndFillBuffer(StripeReader.java:462)
        at org.apache.hadoop.hdfs.StatefulStripeReader.decode(StatefulStripeReader.java:94)
        at org.apache.hadoop.hdfs.StripeReader.readStripe(StripeReader.java:406)
        at org.apache.hadoop.hdfs.DFSStripedInputStream.readOneStripe(DFSStripedInputStream.java:327)
        at org.apache.hadoop.hdfs.DFSStripedInputStream.readWithStrategy(DFSStripedInputStream.java:420)
        at org.apache.hadoop.hdfs.DFSInputStream.read(DFSInputStream.java:892)
        at java.base/java.io.DataInputStream.read(DataInputStream.java:149)
        at java.base/java.io.DataInputStream.read(DataInputStream.java:149) 

 

while (!futures.isEmpty()) {
  try {
    StripingChunkReadResult r = StripedBlockUtil
        .getNextCompletedStripedRead(service, futures, 0);
    dfsStripedInputStream.updateReadStats(r.getReadStats());
    DFSClient.LOG.debug("Read task returned: {}, for stripe {}",
        r, alignedStripe);
    StripingChunk returnedChunk = alignedStripe.chunks[r.index];
    Preconditions.checkNotNull(returnedChunk);
    Preconditions.checkState(returnedChunk.state == StripingChunk.PENDING);

    if (r.state == StripingChunkReadResult.SUCCESSFUL) {
      returnedChunk.state = StripingChunk.FETCHED;
      alignedStripe.fetchedChunksNum++;
      updateState4SuccessRead(r);
      if (alignedStripe.fetchedChunksNum == dataBlkNum) {
        clearFutures();
        break;
      }
    } else {
      returnedChunk.state = StripingChunk.MISSING;
      // close the corresponding reader
      dfsStripedInputStream.closeReader(readerInfos[r.index]);

      final int missing = alignedStripe.missingChunksNum;
      alignedStripe.missingChunksNum++;
      checkMissingBlocks();

      readDataForDecoding();
      readParityChunks(alignedStripe.missingChunksNum - missing);
    } 

This error can be trigger by #StatefulStripeReader.decode.

The reason is that:

1. If there are more than one data block read failed, the #readDataForDecoding will be called multiple times;
2. The decodeInputs array will be initialized repeatedly.
3. The parity data in decodeInputs array which filled by #readParityChunks previously will be set to null.

 

 

 

Issue Links

• links to: GitHub Pull Request #4167