[ 
https://issues.apache.org/jira/browse/HDFS-13511?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16496181#comment-16496181
 ] 

Xiao Chen commented on HDFS-13511:
----------------------------------

Thanks [[email protected]] for creating the jira, and [~gabor.bota] for 
working on this.

Some comments:
 - For downstream to use this, [compat 
guideline|http://hadoop.apache.org/docs/r3.0.0/hadoop-project-dist/hadoop-common/Compatibility.html]
 requires it to be public. Suggest to use Public + Unstable for the initial 
version
 - {{LocatedBlock}} is Private, let's not expose that via this exception. We 
can just construct a string message on the fly, and do not cache the 
LocatedBlock object. This way, we also don't have to worry about whether the 
{{LocatedBlock}} is mutable or not. See {{ReplicaNotFoundException}} for 
example.

> Provide specialized exception when block length cannot be obtained
> ------------------------------------------------------------------
>
>                 Key: HDFS-13511
>                 URL: https://issues.apache.org/jira/browse/HDFS-13511
>             Project: Hadoop HDFS
>          Issue Type: Improvement
>            Reporter: Ted Yu
>            Assignee: Gabor Bota
>            Priority: Major
>         Attachments: HDFS-13511.001.patch
>
>
> In downstream project, I saw the following code:
> {code}
>         FSDataInputStream inputStream = hdfs.open(new Path(path));
> ...
>         if (options.getRecoverFailedOpen() && dfs != null && 
> e.getMessage().toLowerCase()
>             .startsWith("cannot obtain block length for")) {
> {code}
> The above tightly depends on the following in DFSInputStream#readBlockLength
> {code}
>     throw new IOException("Cannot obtain block length for " + locatedblock);
> {code}
> The check based on string matching is brittle in production deployment.
> After discussing with [[email protected]], better approach is to introduce 
> specialized IOException, e.g. CannotObtainBlockLengthException so that 
> downstream project doesn't have to rely on string matching.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

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

Reply via email to