Sharmadha Sainath created ATLAS-1948:
----------------------------------------
Summary: Importing hive_table in a database which is a CTAS of
another table in different database throws exception due to export order.
Key: ATLAS-1948
URL: https://issues.apache.org/jira/browse/ATLAS-1948
Project: Atlas
Issue Type: Bug
Components: atlas-core
Affects Versions: 0.9-incubating
Reporter: Sharmadha Sainath
Priority: Critical
Fix For: 0.9-incubating
Attachments: ImportTransformsErrorOnCTASonDiffDB.txt
1.Created 2 databases db1 , db2 in cluster1
2.Created 2 tables
1. db1.t1
2. db2.t2 as select * from db1.t1
3.Exported db1.t1 into zip file.
4.Imported zip file into cluster 2 with transforms option :
{code}
{
"options": {
"transforms": "{ \"hive_column\": { \"qualifiedName\": [ \"replace:cl1:cl2\"
]} }"
}
}
{code}
5. Import fails with
{code}
{"errorCode":"ATLAS-500-00-001","errorMessage":"org.apache.atlas.exception.AtlasBaseException:
ObjectId is not valid
AtlasObjectId{guid='51c77c1e-265e-46ab-bbb5-5316cf80a53c',
typeName='hive_column', uniqueAttributes={}}"}
{code}
Only db1.t1 is imported into Atlas without any lineage.
Attached the exception stack trace.
After this exporting db2.t2 and importing completes successfully.
That is , first import ,either db1.t1 or db2.t1 is unsuccessful with exception.
Next import is successful.
The exception *doesn't* happen and tables are successfully imported If both the
tables are in a single database. Export order if tables are in same db is
1.table1,
2.db,
3.table2,
4.hive_process
5. hive_column_lineage
If the tables are in different db , the order is ,
1.table1,
2.db1,
3.hive_process,
4.hive_column_lineage
5.ctas table
6.db2
which is possibly causing the issue.
When cluster2 starts importing , it imports table1 , db1 and when it comes to
hive_column_lineage , it finds that column specified in hive_column_lineage is
not in cluster2 yet ,since ctas table comes after the hive_column_lineage in
import order and it throws "ObjectId is not valid
AtlasObjectId{guid='51c77c1e-265e-46ab-bbb5-5316cf80a53c',
typeName='hive_column' ".
Thanks [~ayubkhan] for the analysis.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
