Abin Shahab created HIVE-6670:
---------------------------------
Summary: ClassNotFound with Serde
Key: HIVE-6670
URL: https://issues.apache.org/jira/browse/HIVE-6670
Project: Hive
Issue Type: Bug
Affects Versions: 0.12.0
Reporter: Abin Shahab
We are finding a ClassNotFound exception when we use
CSVSerde(https://github.com/ogrodnek/csv-serde) to create a table.
This is happening because MapredLocalTask does not pass the local added jars to
ExecDriver when that is launched.
ExecDriver's classpath does not include the added jars. Therefore, when the
plan is deserialized, it throws a ClassNotFoundException in the deserialization
code, and results in a TableDesc object with a Null DeserializerClass.
This results in an NPE during Fetch.
Steps to reproduce:
wget
https://drone.io/github.com/ogrodnek/csv-serde/files/target/csv-serde-1.1.2-0.11.0-all.jar
into somewhere local eg.
/home/soam/HiveSerdeIssue/csv-serde-1.1.2-0.11.0-all.jar.
Place the sample files attached to this ticket in HDFS as follows:
hdfs dfs -mkdir /user/soam/HiveSerdeIssue/sampleCSV/
hdfs dfs -put /home/soam/sampleCSV.csv /user/soam/HiveSerdeIssue/sampleCSV/
hdfs dfs -mkdir /user/soam/HiveSerdeIssue/sampleJoinTarget/
hdfs dfs -put /home/soam/sampleJoinTarget.csv
/user/soam/HiveSerdeIssue/sampleJoinTarget/
====
create the tables in hive (this might cause a problem in dogfood since i've
already created tables in those names, so you'll have to change the table names
or delete mine):
ADD JAR /home/soam/HiveSerdeIssue/csv-serde-1.1.2-0.11.0-all.jar;
create external table sampleCSV (md5hash string, filepath string)
row format serde 'com.bizo.hive.serde.csv.CSVSerde'
stored as textfile
location '/user/soam/HiveSerdeIssue/sampleCSV/'
;
create external table sampleJoinTarget (md5hash string, filepath string,
datestamp string, nblines string, nberrors string)
ROW FORMAT DELIMITED
FIELDS TERMINATED BY ','
LINES TERMINATED BY '\n'
STORED AS TEXTFILE
LOCATION '/user/soam/HiveSerdeIssue/sampleJoinTarget/'
;
===============
Now, try the following JOIN:
ADD JAR /home/soam/HiveSerdeIssue/csv-serde-1.1.2-0.11.0-all.jar;
SELECT
sampleCSV.md5hash,
sampleCSV.filepath
FROM sampleCSV
JOIN sampleJoinTarget
ON (sampleCSV.md5hash = sampleJoinTarget.md5hash)
;
—
This will fail with the error:
Execution log at: /tmp/soam/.log
java.lang.ClassNotFoundException: com/bizo/hive/serde/csv/CSVSerde
Continuing ...
2014-03-11 10:35:03 Starting to launch local task to process map join; maximum
memory = 238551040
Execution failed with exit status: 2
Obtaining error information
Task failed!
Task ID:
Stage-4
Logs:
/var/log/hive/soam/hive.log
FAILED: Execution Error, return code 2 from
org.apache.hadoop.hive.ql.exec.mr.MapredLocalTask
Try the following LEFT JOIN. This will work:
SELECT
sampleCSV.md5hash,
sampleCSV.filepath
FROM sampleCSV
LEFT JOIN sampleJoinTarget
ON (sampleCSV.md5hash = sampleJoinTarget.md5hash)
;
==
--
This message was sent by Atlassian JIRA
(v6.2#6252)
