Uh oh!
There was an error while loading. Please reload this page.
[SPARK-19970][SQL][BRANCH-2.1] Table owner should be USER instead of PRINCIPAL in kerberized clusters - #17363
[SPARK-19970][SQL][BRANCH-2.1] Table owner should be USER instead of PRINCIPAL in kerberized clusters#17363dongjoon-hyun wants to merge 1 commit into
Conversation
…PRINCIPAL in kerberized clusters
| } | ||
| hiveTable.setPartCols(partCols.asJava) | ||
| hiveTable.setOwner(conf.getUser) | ||
| hiveTable.setOwner(state.getAuthenticator().getUserName()) |
SparkQA
commented
Mar 20, 2017
Test build #74891 has finished for PR 17363 at commit
|
This is also tested manually in kerberized cluster, @vanzin . BTW, Spark 1.6 has the same issue at ClientWrapper. According to the Spark dev email, there exists demands on more Apache Spark 1.6.X. May I create a backport for branch-1.6? How do you think about that? For Spark 1.6, it happens only with CREATE TABLE .. AS SELECT statement. |
dongjoon-hyun
commented
Mar 22, 2017
Hi, @vanzin . |
dongjoon-hyun
commented
Mar 22, 2017
Retest this please |
SparkQA
commented
Mar 22, 2017
Test build #75055 has finished for PR 17363 at commit
|
vanzin
commented
Mar 23, 2017
Merging to 2.1 / 2.0. I don't really expect any more 1.6 releases at this point, so I wouldn't bother. |
…PRINCIPAL in kerberized clusters ## What changes were proposed in this pull request? In the kerberized hadoop cluster, when Spark creates tables, the owner of tables are filled with PRINCIPAL strings instead of USER names. This is inconsistent with Hive and causes problems when using [ROLE](https://cwiki.apache.org/confluence/display/Hive/SQL+Standard+Based+Hive+Authorization) in Hive. We had better to fix this. **BEFORE** ```scala scala> sql("create table t(a int)").show scala> sql("desc formatted t").show(false) ... |Owner: |sparkEXAMPLE.COM | | ``` **AFTER** ```scala scala> sql("create table t(a int)").show scala> sql("desc formatted t").show(false) ... |Owner: |spark | | ``` ## How was this patch tested? Manually do `create table` and `desc formatted` because this happens in Kerberized clusters. Author: Dongjoon Hyun <dongjoon@apache.org> Closes#17363 from dongjoon-hyun/SPARK-19970-2.
vanzin
commented
Mar 23, 2017
Failed to merge to 2.0. :-/ |
vanzin
commented
Mar 23, 2017
@dongjoon-hyun you'll have to manually close this. |
dongjoon-hyun
commented
Mar 23, 2017
Oo. Thank you so much, @vanzin ! |
What changes were proposed in this pull request?
In the kerberized hadoop cluster, when Spark creates tables, the owner of tables are filled with PRINCIPAL strings instead of USER names. This is inconsistent with Hive and causes problems when using ROLE in Hive. We had better to fix this.
BEFORE
AFTER
How was this patch tested?
Manually do
create tableanddesc formattedbecause this happens in Kerberized clusters.