[
https://issues.apache.org/jira/browse/FLINK-25529?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17491785#comment-17491785
]
luoyuxia edited comment on FLINK-25529 at 2/14/22, 4:08 AM:
------------------------------------------------------------
[~straw] yes, it's class conflict. The OrcWriter in orc-1.5.6 will call method
TimestampColumnVector.isUTC(), but jvm load the class TimestampColumnVector
included in hive-exec, which don't have the method. Two quick way may help to
fix the issue:
1. Replace orc-core-1.5.6 with [orc-core-1.5.6-nohive.jar
|https://repo1.maven.org/maven2/org/apache/orc/orc-core/1.5.6/orc-core-1.5.6-nohive.jar],
then you won't need the hive-storage-api-2.6.0.jar.
2.set table.exec.hive.fallback-mapred-writer = true to avoid using
flink-orc-writer as a workaround. It seems there are some compatibility issues
for hive in flink-orc-writer.
was (Author: luoyuxia):
[~straw] yes, it's class conflicts. The OrcWriter in orc-1.5.6 will call method
TimestampColumnVector.isUTC(), but jvm load the class TimestampColumnVector
included in hive-exec, which don't have the method. Two quick way may help to
fix the issue:
1. Replace orc-core-1.5.6 with [orc-core-1.5.6-nohive.jar
|https://repo1.maven.org/maven2/org/apache/orc/orc-core/1.5.6/orc-core-1.5.6-nohive.jar],
then you won't need the hive-storage-api-2.6.0.jar.
2.set table.exec.hive.fallback-mapred-writer = true to avoid using
flink-orc-writer as a workaround. It seems there are some compatibility issues
for hive in flink-orc-writer.
> java.lang.ClassNotFoundException: org.apache.orc.PhysicalWriter when write
> bulkly into hive-2.1.1 orc table
> -----------------------------------------------------------------------------------------------------------
>
> Key: FLINK-25529
> URL: https://issues.apache.org/jira/browse/FLINK-25529
> Project: Flink
> Issue Type: Bug
> Components: Connectors / Hive
> Environment: hive 2.1.1
> flink 1.12.4
> Reporter: Yuan Zhu
> Priority: Major
> Labels: pull-request-available
> Attachments: lib.jpg
>
>
> I tried to write data bulkly into hive-2.1.1 with orc format, and encountered
> java.lang.ClassNotFoundException: org.apache.orc.PhysicalWriter
>
> Using bulk writer by setting table.exec.hive.fallback-mapred-writer = false;
>
> {code:java}
> SET 'table.sql-dialect'='hive';
> create table orders(
> order_id int,
> order_date timestamp,
> customer_name string,
> price decimal(10,3),
> product_id int,
> order_status boolean
> )partitioned by (dt string)
> stored as orc;
>
> SET 'table.sql-dialect'='default';
> create table datagen_source (
> order_id int,
> order_date timestamp(9),
> customer_name varchar,
> price decimal(10,3),
> product_id int,
> order_status boolean
> )with('connector' = 'datagen');
> create catalog myhive with ('type' = 'hive', 'hive-conf-dir' = '/mnt/conf');
> set table.exec.hive.fallback-mapred-writer = false;
> insert into myhive.`default`.orders
> /*+ OPTIONS(
> 'sink.partition-commit.trigger'='process-time',
> 'sink.partition-commit.policy.kind'='metastore,success-file',
> 'sink.rolling-policy.file-size'='128MB',
> 'sink.rolling-policy.rollover-interval'='10s',
> 'sink.rolling-policy.check-interval'='10s',
> 'auto-compaction'='true',
> 'compaction.file-size'='1MB' ) */
> select * , date_format(now(),'yyyy-MM-dd') as dt from datagen_source; {code}
> [ERROR] Could not execute SQL statement. Reason:
> java.lang.ClassNotFoundException: org.apache.orc.PhysicalWriter
>
> My jars in lib dir listed in attachment.
> In HiveTableSink#createStreamSink(line:270), createBulkWriterFactory if
> table.exec.hive.fallback-mapred-writer is false.
> If table is orc, HiveShimV200#createOrcBulkWriterFactory will be invoked.
> OrcBulkWriterFactory depends on org.apache.orc.PhysicalWriter in orc-core,
> but flink-connector-hive excludes orc-core for conflicting with hive-exec.
>
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
