[AMBARI-22444] Add Native Libraries To Tez Tarball

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 2.6.1
• Fix Version/s: 2.6.1
• Assignee: Jonathan Hurley
• Reporter: Jonathan Hurley
• Created: 14/Nov/17 21:17
• Updated: 16/Nov/17 21:18
• Resolved: 15/Nov/17 19:56

Description

As part of the investigation for getting patch upgrades working in Ambari 2.6.1, it was determined that the Tez tarball will need to have the hadoop native libraries added to it so that they can be detected from the tarball.

STR:

• Install ZK, MapR, Tez, Yarn, Hive
• Enable a non-LZO codec, like Snappy
• Patch Hive to a new version
• Change the following properties in tez-site:
  • tez.am.launch.env = LD_LIBRARY=./tezlib/lib/native
  • tez.task.launch.env = LD_LIBRARY=./tezlib/lib/native

When Hive commands run, they will attempt to load the native snappy libraries from the Tez tarball and will fail with:

Caused by: java.io.IOException: Unable to get CompressorType for codec (org.apache.hadoop.io.compress.SnappyCodec). This is most likely due to missing native libraries for the codec.
	at org.apache.tez.runtime.library.common.sort.impl.ExternalSorter.<init>(ExternalSorter.java:217)

In order to fix this, the Tez tarball should include the native hadoop libraries as well:

├── tez
│   ├── lib
│   │   ├── native
│   │   │   ├── libhadoop.a
│   │   │   ├── libhadoop.so -> libhadoop.so.1.0.0
│   │   │   ├── libhadoop.so.1.0.0
│   │   │   ├── libhadooppipes.a
│   │   │   ├── libhadooputils.a
│   │   │   ├── libhdfs.a
│   │   │   ├── libhdfs.so -> libhdfs.so.0.0.0
│   │   │   ├── libhdfs.so.0.0.0
│   │   │   ├── libsnappy.so -> libsnappy.so.1.1.4
│   │   │   ├── libsnappy.so.1 -> libsnappy.so.1.1.4
│   │   │   └── libsnappy.so.1.1.4

Issue Links

• links to: ReviewBoard