[
https://issues.apache.org/jira/browse/DRILL-6726?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16756501#comment-16756501
]
Robert Hou commented on DRILL-6726:
-----------------------------------
I have encountered another problem related to this one. If I run Drill 1.15,
and then I run Drill 1.14, Drill 1.14 cannot access schemas using mixed-case
(have upper case letters). It can access the schema if it uses lower case
letters. For example, if the schema used to be called "drillTestDir", Drill
1.14 must use "drilltestdir" in order to use it. This means that scripts that
use "drillTestDir" can break.
This may not be a major issue now, but sometimes users can try a new version of
Drill, and if they run into problems, they can revert to the older version of
Drill. We know one user who tried Drill 1.14 and encountered some problems and
went back to Drill 1.13. We should keep this in mind in future releases.
> Drill fails to query views created before DRILL-6492 when impersonation is
> enabled
> ----------------------------------------------------------------------------------
>
> Key: DRILL-6726
> URL: https://issues.apache.org/jira/browse/DRILL-6726
> Project: Apache Drill
> Issue Type: Bug
> Components: Execution - Relational Operators
> Affects Versions: 1.15.0
> Reporter: Robert Hou
> Assignee: Arina Ielchiieva
> Priority: Major
> Labels: ready-to-commit
> Fix For: 1.15.0
>
> Attachments: student
>
>
> Drill 1.14 changes schemas to be case-insensitive (DRILL-6492). If an
> existing view was created before (DRILL-6492) was committed, and this view
> references a file that includes a schema which has upper case letters, the
> view needs to be rebuilt. There may be variations on this issue that I have
> not seen.
> To reproduce this problem, create a dfs workspace like this:
> {noformat}
> "drillTestDirP1": {
> "location": "/drill/testdata/p1tests",
> "writable": true,
> "defaultInputFormat": "parquet",
> "allowAccessOutsideWorkspace": false
> },
> {noformat}
> Use Drill commit ddb35ce71837376c7caef28c25327ba556bb32f2 and execute this
> command:
> {noformat}
> create or replace view `dfs.drillTestDirP1`.student_parquet_v as select *
> from `dfs.drillTestDirP1`.student;
> {noformat}
> Then use Drill commit ddb35ce71837376c7caef28c25327ba556bb32f2 and execute
> this query:
> {noformat}
> select * from student_test_v;
> {noformat}
> Drill will return an exception:
> {noformat}
> Error: VALIDATION ERROR: Failure while attempting to expand view. Requested
> schema drillTestDirP1 not available in schema dfs.
> View Context dfs, drillTestDirP1
> View SQL SELECT *
> FROM `dfs.drillTestDirP1`.`student`
> [Error Id: 3f4594ee-b503-40db-8845-474b0ecb5feb on qa-node211.qa.lab:31010]
> (state=,code=0)
> {noformat}
> I have attached the student parquet file I used.
> This is what the .view.drill file looks like:
> {noformat}
> {
> "name" : "student_test_v",
> "sql" : "SELECT *\nFROM `dfs.drillTestDirP1`.`student`",
> "fields" : [ {
> "name" : "**",
> "type" : "DYNAMIC_STAR",
> "isNullable" : true
> } ],
> "workspaceSchemaPath" : [ "dfs", "drillTestDirP1" ]
> }
> {noformat}
> This means that users may not be able to access views that they have created
> using previous versions of Drill. We should maintain backwards
> compatibiliity where possible.
> As work-around, these views can be re-created. It would be helpful to users
> if the error message explains that these views need to be re-created.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
