[
https://issues.apache.org/jira/browse/HDFS-10587?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15374480#comment-15374480
]
Yongjun Zhang commented on HDFS-10587:
--------------------------------------
About the visibleLength, I saw
{code}
In ReplicaBeingWritten.java
@Override
public long getVisibleLength() {
return getBytesAcked(); // all acked bytes are visible
}
{code}
which means different replicas may have different visibleLength, because
BytesAcked at different DataNodes maybe different.
My earlier effort was to claim that using different visibleLength at the
BlockReceiver than the BlockSender side is wrong. Based on the above code, it
might be ok to claim the visibleLength as the received data length at the
destination side of blockTransfer (better to get confirmation though).
So, we need to understand, how the corruption really happened, and where in the
block data: Did it happen when we receive this chunk of data, or when we
receive new data after reconstructing the pipeline? Because based on my
analysis so far, the skipping of the bytes on disk (mentioned in the following
statement) is necessary since the data is not garbage (assuming the data at the
Sender side is good).
{quote}
(8) When new data was appended to the destination, it skipped the bytes already
on disk. Therefore, whatever was written as garbage was not replaced.
{quote}
One possibility is that the checksum handling there is not correct in a corner
situation.
If we have a testcase to replicate the issue, we need to look at both the
source side data and destination side data, to see whether it's real data
corruption, or checksum miscalculation. If there is corruption, where exactly
the corruption is.
> Incorrect offset/length calculation in pipeline recovery causes block
> corruption
> --------------------------------------------------------------------------------
>
> Key: HDFS-10587
> URL: https://issues.apache.org/jira/browse/HDFS-10587
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode
> Reporter: Wei-Chiu Chuang
> Assignee: Wei-Chiu Chuang
> Attachments: HDFS-10587.001.patch
>
>
> We found incorrect offset and length calculation in pipeline recovery may
> cause block corruption and results in missing blocks under a very unfortunate
> scenario.
> (1) A client established pipeline and started writing data to the pipeline.
> (2) One of the data node in the pipeline restarted, closing the socket, and
> some written data were unacknowledged.
> (3) Client replaced the failed data node with a new one, initiating block
> transfer to copy existing data in the block to the new datanode.
> (4) The block is transferred to the new node. Crucially, the entire block,
> including the unacknowledged data, was transferred.
> (5) The last chunk (512 bytes) was not a full chunk, but the destination
> still reserved the whole chunk in its buffer, and wrote the entire buffer to
> disk, therefore some written data is garbage.
> (6) When the transfer was done, the destination data node converted the
> replica from temporary to rbw, which made its visible length as the length of
> bytes on disk. That is to say, it thought whatever was transferred was
> acknowledged. However, the visible length of the replica is different (round
> up to the next multiple of 512) than the source of transfer. [1]
> (7) Client then truncated the block in the attempt to remove unacknowledged
> data. However, because the visible length is equivalent of the bytes on disk,
> it did not truncate unacknowledged data.
> (8) When new data was appended to the destination, it skipped the bytes
> already on disk. Therefore, whatever was written as garbage was not replaced.
> (9) the volume scanner detected corrupt replica, but due to HDFS-10512, it
> wouldn’t tell NameNode to mark the replica as corrupt, so the client
> continued to form a pipeline using the corrupt replica.
> (10) Finally the DN that had the only healthy replica was restarted. NameNode
> then update the pipeline to only contain the corrupt replica.
> (11) Client continue to write to the corrupt replica, because neither client
> nor the data node itself knows the replica is corrupt. When the restarted
> datanodes comes back, their replica are stale, despite they are not corrupt.
> Therefore, none of the replica is good and up to date.
> The sequence of events was reconstructed based on DataNode/NameNode log and
> my understanding of code.
> Incidentally, we have observed the same sequence of events on two independent
> clusters.
> [1]
> The sender has the replica as follows:
> 2016-04-15 22:03:05,066 INFO
> org.apache.hadoop.hdfs.server.datanode.fsdataset.impl.FsDatasetImpl:
> Recovering ReplicaBeingWritten, blk_1556997324_1100153495099, RBW
> getNumBytes() = 41381376
> getBytesOnDisk() = 41381376
> getVisibleLength()= 41186444
> getVolume() = /hadoop-i/data/current
> getBlockFile() =
> /hadoop-i/data/current/BP-1043567091-10.216.26.120-1343682168507/current/rbw/blk_1556997324
> bytesAcked=41186444
> bytesOnDisk=41381376
> while the receiver has the replica as follows:
> 2016-04-15 22:03:05,068 INFO
> org.apache.hadoop.hdfs.server.datanode.fsdataset.impl.FsDatasetImpl:
> Recovering ReplicaBeingWritten, blk_1556997324_1100153495099, RBW
> getNumBytes() = 41186816
> getBytesOnDisk() = 41186816
> getVisibleLength()= 41186816
> getVolume() = /hadoop-g/data/current
> getBlockFile() =
> /hadoop-g/data/current/BP-1043567091-10.216.26.120-1343682168507/current/rbw/blk_1556997324
> bytesAcked=41186816
> bytesOnDisk=41186816
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
