[ 
https://issues.apache.org/jira/browse/HDFS-11644?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15966723#comment-15966723
 ] 

Andrew Wang commented on HDFS-11644:
------------------------------------

Looking at this, it's not pretty. FileSystem returns an FSDataOutputStream, 
which implements Syncable. Its implementation either does a real hflush, or 
just calls flush. See:

{code:title=FSDataOutputStream}
  @Override  // Syncable
  public void hflush() throws IOException {
    if (wrappedStream instanceof Syncable) {
      ((Syncable)wrappedStream).hflush();
    } else {
      wrappedStream.flush();
    }
  }
{code}

I don't understand how users can figure out if they're getting a real hflush. 
FSDataOutputStream implements Syncable, so you can't query with {{instanceof}}. 
There's currently no public way of querying the wrapped stream either. I think 
it was a mistake to add {{Syncable}} to FSDataOutputStream, we should have 
forced users to check with {{instanceof}} and cast it.

I don't like changing DFSStripedOutputStream#hflush to simply call flush, since 
then HDFS users who turn on EC will silently stop getting real hflush/hsync. 
The current behavior of throwing an exception is safer.

[[email protected]], any thoughts on this? I notice that output streams aren't 
covered by the FileSystem spec. This also relates to discussions about querying 
which features are supported by a FS.

> DFSStripedOutputStream should not implement Syncable
> ----------------------------------------------------
>
>                 Key: HDFS-11644
>                 URL: https://issues.apache.org/jira/browse/HDFS-11644
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: erasure-coding
>    Affects Versions: 3.0.0-alpha1
>            Reporter: Andrew Wang
>            Assignee: Manoj Govindassamy
>              Labels: hdfs-ec-3.0-must-do
>
> FSDataOutputStream#hsync checks if a stream implements Syncable, and if so, 
> calls hsync. Otherwise, it just calls flush. This is used, for instance, by 
> YARN's FileSystemTimelineWriter.
> DFSStripedOutputStream extends DFSOutputStream, which implements Syncable. 
> However, DFSStripedOS throws a runtime exception when the Syncable methods 
> are called.
> We should refactor the inheritance structure so DFSStripedOS does not 
> implement Syncable.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to