Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
HDFS >> mail # dev >> Re: data loss after cluster wide power loss

Copy link to this message
Re: data loss after cluster wide power loss

Thanks for the detailed email. Sorry I did not read all the details you had
sent earlier completely (on my phone). As you said, this is not related to
data loss related to HBase log and hsync. I think you are right; the rename
operation itself might not have hit the disk. I think we should either
ensure metadata operation is synced on the datanode or handle it being
reported as blockBeingWritten. Let me spend sometime to debug this issue.

One surprising thing is, all the replicas were reported as

On Mon, Jul 1, 2013 at 6:03 PM, Dave Latham <[EMAIL PROTECTED]> wrote:

> (Removing hbase list and adding hdfs-dev list as this is pretty internal
> stuff).
> Reading through the code a bit:
> FSDataOutputStream.close calls
> DFSOutputStream.close calls
> DFSOutputStream.closeInternal
>  - sets currentPacket.lastPacketInBlock = true
>  - then calls
> DFSOutputStream.flushInternal
>  - enqueues current packet
>  - waits for ack
> BlockReceiver.run
>  - if (lastPacketInBlock && !receiver.finalized) calls
> FSDataset.finalizeBlock calls
> FSDataset.finalizeBlockInternal calls
> FSVolume.addBlock calls
> FSDir.addBlock calls
> FSDir.addBlock
>  - renames block from "blocksBeingWritten" tmp dir to "current" dest dir
> This looks to me as I would expect a synchronous chain from a DFS client
> to moving the file from blocksBeingWritten to the current dir so that once
> the file is closed that it the block files would be in the proper directory
> - even if the contents of the file are still in the OS buffer rather than
> synced to disk.  It's only after this moving of blocks that
> NameNode.complete file is called.  There are several conditions and loops
> in there that I'm not certain this chain is fully reliable in all cases
> without a greater understanding of the code.
> Could it be the case that the rename operation itself is not synced and
> that ext3 lost the fact that the block files were moved?
> Or is there a bug in the close file logic that for some reason the block
> files are not always moved into place when a file is closed?
> Thanks for your patience,
> Dave
> On Mon, Jul 1, 2013 at 3:35 PM, Dave Latham <[EMAIL PROTECTED]> wrote:
>> Thanks for the response, Suresh.
>> I'm not sure that I understand the details properly.  From my reading of
>> HDFS-744 the hsync API would allow a client to make sure that at any point
>> in time it's writes so far hit the disk.  For example, for HBase it could
>> apply a fsync after adding some edits to its WAL to ensure those edits are
>> fully durable for a file which is still open.
>> However, in this case the dfs file was closed and even renamed.  Is it
>> the case that even after a dfs file is closed and renamed that the data
>> blocks would still not be synced and would still be stored by the datanode
>> in "blocksBeingWritten" rather than in "current"?  If that is case, would
>> it be better for the NameNode not to reject replicas that are in
>> blocksBeingWritten, especially if it doesn't have any other replicas
>> available?
>> Dave
>> On Mon, Jul 1, 2013 at 3:16 PM, Suresh Srinivas <[EMAIL PROTECTED]>wrote:
>>> Yes this is a known issue.
>>> The HDFS part of this was addressed in
>>> https://issues.apache.org/jira/browse/HDFS-744 for 2.0.2-alpha and is
>>> not
>>> available in 1.x  release. I think HBase does not use this API yet.
>>> On Mon, Jul 1, 2013 at 3:00 PM, Dave Latham <[EMAIL PROTECTED]> wrote:
>>> > We're running HBase over HDFS 1.0.2 on about 1000 nodes.  On Saturday
>>> the
>>> > data center we were in had a total power failure and the cluster went
>>> down
>>> > hard.  When we brought it back up, HDFS reported 4 files as CORRUPT.
>>>  We
>>> > recovered the data in question from our secondary datacenter, but I'm
>>> > trying to understand what happened and whether this is a bug in HDFS
>>> that
>>> > should be fixed.
>>> >
>>> > From what I can tell the file was created and closed by the dfs client