hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sailesh Mukil (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HDFS-11529) libHDFS still does not return appropriate error information in many cases
Date Tue, 14 Mar 2017 09:08:41 GMT

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

Sailesh Mukil updated HDFS-11529:
---------------------------------
    Description: 
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.


  was:
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. 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.



> libHDFS still does not return appropriate error information in many cases
> -------------------------------------------------------------------------
>
>                 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
>            Priority: Critical
>              Labels: errorhandling, libhdfs
>
> 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


Mime
View raw message