[ https://issues.apache.org/jira/browse/HDFS-11529?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16172067#comment-16172067 ]
John Zhuge commented on HDFS-11529: ----------------------------------- The patch has a bug tracked by HDFS-12494 libhdfs SIGSEGV in setTLSExceptionStrings. > 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 > Fix For: 3.0.0-alpha4 > > Attachments: HDFS-11529.000.patch, HDFS-11529.001.patch, > HDFS-11529.002.patch, HDFS-11529.003.patch, HDFS-11529.004.patch, > HDFS-11529.005.patch, HDFS-11529.006.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 (v6.4.14#64029) --------------------------------------------------------------------- To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org