zhuwei created HIVE-20725:
-----------------------------
Summary: Simultaneous dynamic inserts can result in partition
files lost
Key: HIVE-20725
URL: https://issues.apache.org/jira/browse/HIVE-20725
Project: Hive
Issue Type: Bug
Reporter: zhuwei
Assignee: zhuwei
If two users attempt a dynamic insert into the same new partition at the same
time, a possible race condition exists which result in error state. In that
case the partition info has been inserted to metastore but data files been
removed.
The current logic in function "add_partition_core" in class
HiveMetaStore.HMSHandler is like this :
# check if partition already exists
# create the partition files directory if not exists
# try to add partition
# if add partition failed and it created the directory in step 2, delete that
directory
Assume that two users are inserting the same partition at the same time, there
are two threads operating their requests, say thread A and thread B. If 1~4
steps of thread B are all done between step 2 and step 3 of thread A. The
sequence like this : A1 A2 B1 B2 B3 B4 A3 A4. The partition files written by B
will be removed by A.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
