[jira] [Updated] (HDFS-11529) Add libHDFS API to return last exception

["John Zhuge (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/HDFS-11529?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

John Zhuge updated HDFS-11529:
------------------------------
    Summary: Add libHDFS API to return last exception  (was: libHDFS still does not return appropriate error information in many cases)

> Add libHDFS API to return last exception
> ----------------------------------------
>
>                 Key: HDFS-11529
>                 URL: https://issues.apache.org/jira/browse/HDFS-11529
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: libhdfs
>    Affects Versions: 2.6.0
>            Reporter: Sailesh Mukil
>            Assignee: Sailesh Mukil
>            Priority: Critical
>              Labels: errorhandling, libhdfs
>         Attachments: HDFS-11529.000.patch, HDFS-11529.001.patch, HDFS-11529.002.patch
>
>
> libHDFS uses a table to compare exceptions against and returns a corresponding error code to the application in case of an error.
> However, this table is manually populated and many times is disremembered when new exceptions are added.
> This causes libHDFS to return EINTERNAL (or Unknown Error(255)) whenever these exceptions are hit. These are some examples of exceptions that have been observed on an Error(255):
> org.apache.hadoop.ipc.StandbyException (Operation category WRITE is not supported in state standby)
> java.io.EOFException: Cannot seek after EOF
> javax.security.sasl.SaslException: GSS initiate failed [Caused by GSSException: No valid credentials provided (Mechanism level: Failed to find any Kerberos tgt)
> It is of course not possible to have an error code for each and every type of exception, so one suggestion of how this can be addressed is by having a call such as hdfsGetLastException() that would return the last exception that a libHDFS thread encountered. This way, an application may choose to call hdfsGetLastException() if it receives EINTERNAL.
> We can make use of the Thread Local Storage to store this information. Also, this makes sure that the current functionality is preserved.
> This is a follow up from HDFS-4997.



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

---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-help@hadoop.apache.org