hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Brian Bockelman (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-4541) Infinite loop in error handler for libhdfs
Date Thu, 30 Oct 2008 20:52:44 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-4541?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12644100#action_12644100
] 

Brian Bockelman commented on HADOOP-4541:
-----------------------------------------

Dummy me forgot the greatest tool for debugging of them all: gdb.

Here's the stack trace of one of these suckers in action:

(gdb) where
#0  0x0000002a975a0a47 in CollectedHeap::allocate_from_tlab_slow () from /opt/osg/osg-100/jdk1.5/jre/lib/amd64/server/libjvm.so
#1  0x0000002a97529551 in CollectedHeap::common_mem_allocate_noinit () from /opt/osg/osg-100/jdk1.5/jre/lib/amd64/server/libjvm.so
#2  0x0000002a9793374b in typeArrayKlass::allocate () from /opt/osg/osg-100/jdk1.5/jre/lib/amd64/server/libjvm.so
#3  0x0000002a9769f5e8 in jni_NewByteArray () from /opt/osg/osg-100/jdk1.5/jre/lib/amd64/server/libjvm.so
#4  0x0000002a971ef363 in hdfsWrite (fs=Variable "fs" is not available.
) at hdfs.c:590
#5  0x0000002a970ea97a in globus_l_gfs_hdfs_dump_buffers (hdfs_handle=0x6de080) at globus_gridftp_server_hdfs.c:507

That dump_buffers call (internal to my application) worries me; give me a few hours to step
through things in GDB; if errors aren't propogated right in that particular function, it could
lead to an application-side infinite loop.

> Infinite loop in error handler for libhdfs
> ------------------------------------------
>
>                 Key: HADOOP-4541
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4541
>             Project: Hadoop Core
>          Issue Type: Bug
>          Components: libhdfs
>    Affects Versions: 0.18.1
>            Reporter: Brian Bockelman
>         Attachments: libhdfs_failure.txt
>
>
> If there is a problem writing out to HDFS, libhdfs gets put in an infinite loop.
> Unfortunately, my program reroutes stderr/stdout to /dev/null, so I am attaching the
strace output.  You can see the java stack traces which are written over and over.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message