Uploaded image for project: 'HBase'
  1. HBase
  2. HBASE-21018

RS crashed because AsyncFS was unable to update HDFS data encryption key

    Details

    • Type: Bug
    • Status: Resolved
    • Priority: Critical
    • Resolution: Fixed
    • Affects Version/s: 2.0.0
    • Fix Version/s: 3.0.0, 2.1.1, 2.0.2
    • Component/s: wal
    • Labels:
      None
    • Environment:

      Hadoop 3.0.0, HBase 2.0.0,
      HDFS configuration dfs.encrypt.data.transfer = true

    • Hadoop Flags:
      Reviewed

      Description

      We (+Umesh Agashe) found HBase RegionServer doesn't update HDFS data encryption key correctly, and in some cases after retry 10 times, it aborts.

      2018-08-03 17:37:03,233 WARN org.apache.hadoop.hbase.io.asyncfs.FanOutOneBlockAsyncDFSOutputHelper: create fan-out dfs output /hbase/WALs/rs1.example.com,22101,1533318719239/rs1.example.com%2C22101%2C1533318719239.rs1.example.com%2C22101%2C1533318719239.regiongroup-0.1533343022981 failed, retry = 1
      org.apache.hadoop.hdfs.protocol.datatransfer.InvalidEncryptionKeyException: Can't re-compute encryption key for nonce, since the required block key (keyID=1685436998) doesn't exist. Current key: 1085959374
              at org.apache.hadoop.hbase.io.asyncfs.FanOutOneBlockAsyncDFSOutputSaslHelper$SaslNegotiateHandler.check(FanOutOneBlockAsyncDFSOutputSaslHelper.java:399)
              at org.apache.hadoop.hbase.io.asyncfs.FanOutOneBlockAsyncDFSOutputSaslHelper$SaslNegotiateHandler.channelRead(FanOutOneBlockAsyncDFSOutputSaslHelper.java:470)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:362)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:348)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:340)
              at org.apache.hbase.thirdparty.io.netty.handler.codec.MessageToMessageDecoder.channelRead(MessageToMessageDecoder.java:102)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:362)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:348)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:340)
              at org.apache.hbase.thirdparty.io.netty.handler.codec.ByteToMessageDecoder.fireChannelRead(ByteToMessageDecoder.java:310)
              at org.apache.hbase.thirdparty.io.netty.handler.codec.ByteToMessageDecoder.channelRead(ByteToMessageDecoder.java:284)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:362)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:348)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:340)
              at org.apache.hbase.thirdparty.io.netty.handler.timeout.IdleStateHandler.channelRead(IdleStateHandler.java:286)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:362)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:348)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.fireChannelRead(AbstractChannelHandlerContext.java:340)
              at org.apache.hbase.thirdparty.io.netty.channel.DefaultChannelPipeline$HeadContext.channelRead(DefaultChannelPipeline.java:1359)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:362)
              at org.apache.hbase.thirdparty.io.netty.channel.AbstractChannelHandlerContext.invokeChannelRead(AbstractChannelHandlerContext.java:348)
              at org.apache.hbase.thirdparty.io.netty.channel.DefaultChannelPipeline.fireChannelRead(DefaultChannelPipeline.java:935)
              at org.apache.hbase.thirdparty.io.netty.channel.epoll.AbstractEpollStreamChannel$EpollStreamUnsafe.epollInReady(AbstractEpollStreamChannel.java:801)
              at org.apache.hbase.thirdparty.io.netty.channel.epoll.EpollEventLoop.processReady(EpollEventLoop.java:404)
              at org.apache.hbase.thirdparty.io.netty.channel.epoll.EpollEventLoop.run(EpollEventLoop.java:304)
              at org.apache.hbase.thirdparty.io.netty.util.concurrent.SingleThreadEventExecutor$5.run(SingleThreadEventExecutor.java:858)
              at org.apache.hbase.thirdparty.io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:138)
              at java.lang.Thread.run(Thread.java:748)
      

      It appears similar to HDFS-10609 and HDFS-11741. There are cases where dfsclient has to clear its local cache of data encryption key when a DataNode finds the key expires. Now that asyncfs implements the fan-out block writer, HBase has to take care of the update of DEK too.

       

      I have a simple fix (similar to what's implemented in HDFS-10609 and HDFS-11741), but making a test seems quite involved.

        Attachments

          Activity

            People

            • Assignee:
              jojochuang Wei-Chiu Chuang
              Reporter:
              jojochuang Wei-Chiu Chuang
            • Votes:
              0 Vote for this issue
              Watchers:
              7 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved: