[
https://issues.apache.org/jira/browse/HIVE-7483?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14714475#comment-14714475
]
Furcy Pin commented on HIVE-7483:
---------------------------------
I confirm, we are having the same issue on Hive 1.1.0, and I reproduced it on
1.2.1
CREATE TABLE test_db.test_table (id INT)
PARTITIONED BY (part STRING)
STORED AS ORC ;
INSERT INTO TABLE test_db.test_table PARTITION (part="test")
VALUES (1), (2), (3), (4)
;
SET hive.exec.dynamic.partition.mode=nonstrict ;
SET hive.support.concurrency=true ;
INSERT OVERWRITE TABLE test_db.test_table PARTITION (part)
SELECT * FROM test_db.test_table ;
nothing happens, and when doing a SHOW LOCKS in another shell we get :
+----------+-----------+------------+------------+-------------+--------------+-----------------+-----------------+----------------+
| lockid | database | table | partition | lock_state | lock_type |
transaction_id | last_heartbeat | acquired_at |
+----------+-----------+------------+------------+-------------+--------------+-----------------+-----------------+----------------+
| 3765 | test_db | test_table | NULL | WAITING | EXCLUSIVE |
NULL | 1440603633148 | NULL |
| 3765 | test_db | test_table | part=test | WAITING | SHARED_READ |
NULL | 1440603633148 | NULL |
+----------+-----------+------------+------------+-------------+--------------+-----------------+-----------------+----------------+
from looking at the source of
org.apache.hadoop.hive.ql.lockmgr.EmbeddedLockManager, I would say it is stuck
in the for loop of the method lock(List<HiveLockObj> objs, int
numRetriesForLock, long sleepTime),
where lockPrimitive() keeps failing on the second lock.
> hive insert overwrite table select from self dead lock
> ------------------------------------------------------
>
> Key: HIVE-7483
> URL: https://issues.apache.org/jira/browse/HIVE-7483
> Project: Hive
> Issue Type: Bug
> Components: Locking
> Affects Versions: 0.13.1
> Reporter: Xiaoyu Wang
>
> CREATE TABLE test(
> id int,
> msg string)
> PARTITIONED BY (
> continent string,
> country string)
> CLUSTERED BY (id)
> INTO 10 BUCKETS
> STORED AS ORC;
> alter table test add partition(continent='Asia',country='India');
> in hive-site.xml:
> hive.txn.manager=org.apache.hadoop.hive.ql.lockmgr.DbTxnManager;
> hive.support.concurrency=true;
> in hive shell:
> set hive.input.format=org.apache.hadoop.hive.ql.io.HiveInputFormat;
> insert into test table some records first.
> then execute sql:
> insert overwrite table test partition(continent='Asia',country='India')
> select id,msg from test;
> the log stop at :
> INFO log.PerfLogger: <PERFLOG method=acquireReadWriteLocks
> from=org.apache.hadoop.hive.ql.Driver>
> i think it has dead lock when insert overwrite table from it self.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
