You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@carbondata.apache.org by "xuchuanyin (JIRA)" <ji...@apache.org> on 2017/12/08 03:46:01 UTC

[jira] [Created] (CARBONDATA-1878) JVM crash after off-heap-sort disabled

xuchuanyin created CARBONDATA-1878:
--------------------------------------

             Summary: JVM crash after off-heap-sort disabled
                 Key: CARBONDATA-1878
                 URL: https://issues.apache.org/jira/browse/CARBONDATA-1878
             Project: CarbonData
          Issue Type: Bug
          Components: core
    Affects Versions: 1.3.0
            Reporter: xuchuanyin
            Assignee: xuchuanyin


# SCENARIO

Recently I have fixed some issues in Carbondata. To perform a full test to cover all the code that has been modified by me, I performed some iteration of the whole test case. Each iteration is started with different  key configurations that will affect the flow in the code.

After I set `enable.offheap.sort=false` (default value is true) in the configuration, running tests will always end up with JVM crash error. The error messages are shown as below:

```
#
# A fatal error has been detected by the Java Runtime Environment:
#
#  SIGSEGV (0xb) at pc=0x00007f346b207ff1, pid=144619, tid=0x00007f346c2fc700
#
# JRE version: Java(TM) SE Runtime Environment (8.0_111-b14) (build 1.8.0_111-b14)
# Java VM: Java HotSpot(TM) 64-Bit Server VM (25.111-b14 mixed mode linux-amd64 compressed oops)
# Problematic frame:
# V  [libjvm.so+0xa90ff1]  Unsafe_SetNativeShort+0x51
#
# Failed to write core dump. Core dumps have been disabled. To enable core dumping, try "ulimit -c unlimited" before starting Java again
#
# An error report file with more information is saved as:
# /home/xu/ws/carbondata/hs_err_pid144619.log
#
# If you would like to submit a bug report, please visit:
#   http://bugreport.java.com/bugreport/crash.jsp
#

Process finished with exit code 134 (interrupted by signal 6: SIGABRT)
```

# STEPS TOREPRODUCE

The error can be easily reproduced in different ways. Here I will provide a simple way to reproduce it:

1. Find the test case `DateDataTypeDirectDictionaryTest`.

2. Add the following code in the method `beforeAll`.
```
CarbonProperties.getInstance()
    .addProperty(CarbonCommonConstants.ENABLE_OFFHEAP_SORT, "false")
```

3. Run this test case.

4. You will find the test failed with the above error.

5. Replace the code in Step2 with the following code:
```
CarbonProperties.getInstance()
    .addProperty(CarbonCommonConstants.ENABLE_OFFHEAP_SORT, "true")
```

6. Run this test case.

7. The test is success without error.

# ANALYZE & RESOLVE

I have reproduced this error and analyzed the core dump file. The final stack message in core dump looks like below:

```
Thread 73303: (state = IN_VM)
 - sun.misc.Unsafe.putShort(long, short) @bci=0 (Interpreted frame)
 - org.apache.carbondata.core.indexstore.UnsafeMemoryDMStore.addToUnsafe(org.apache.carbondata.core.indexstore.schema.CarbonRowSchema, org.apache.carbondata.core.indexstore.row.DataMapRow, int) @bci=781, line=150 (Interpreted frame)
 - org.apache.carbondata.core.indexstore.UnsafeMemoryDMStore.addIndexRowToUnsafe(org.apache.carbondata.core.indexstore.row.DataMapRow) @bci=59, line=99 (Interpreted frame)
 ...
```

After inspecting the code, I found there lies bug in `UnsafeMemoryDMStore line=150` while writing length to unsafe memory -- It writes with wrong base object.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)