[ 
https://issues.apache.org/jira/browse/HDFS-3721?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13422575#comment-13422575
 ] 

Todd Lipcon commented on HDFS-3721:
-----------------------------------

bq. Todd, I will review this in a couple of days. Looking at the issue briefly, 
this is only an issue with DataTransferProtocol, due to packet header length. 
It should not happen in RPC right? I would be good to review RPC code once more 
to confirm it.

I think RPC is basically sound. I've done diffs of the protobufs between trunk 
and 2.0 and all the changes look compatible. This was just a special case 
because, when I did the original PB-ification, I took the simpler route of 
maintaining fixed-length headers. Then I overlooked the bit of HDFS-744 which 
made the header non-fixed-width (oops).

bq. In a related note, we should think about what backward/forward 
compatibility for an alpha release means and how to handle it when we find 
issues.

In that case, I'd also like to consider whether we can promote the HDFS side of 
2.0.x to "beta". We've seen very few issues after extensive testing. I don't 
want the alpha state of MR2 to hold back stability classification of HDFS, 
which is important for those users who aren't using MR2. But that's probably 
better suited for a mailing list discussion.
                
> hsync support broke wire compatibility
> --------------------------------------
>
>                 Key: HDFS-3721
>                 URL: https://issues.apache.org/jira/browse/HDFS-3721
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: data-node, hdfs client
>    Affects Versions: 2.1.0-alpha
>            Reporter: Todd Lipcon
>            Assignee: Todd Lipcon
>            Priority: Critical
>         Attachments: hdfs-3721.txt
>
>
> HDFS-744 added support for hsync to the data transfer wire protocol. However, 
> it actually broke wire compatibility: if the client has hsync support but the 
> server does not, the client cannot read or write data on the old cluster.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: 
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to