Bug Report: Schema Discovery Problem with nested documents (Dremio 3.0.0 and MongoDB 3.6.8)

1

Hi Dremio Team,

I’m facing an issue with Schema Discovery (Preview works, Run broken) when using nested JSON structures in MongoDB. The case is pretty easy to reproduce. I add three JSON documents to a Mongo collection:

[
{
  "o" : {                                   
        "text": "a",       
        "num" : 42                             
      }
    },     
      { 
    "o":
      {                                   
        "text": "b",       
        "num": 43                             
      }
    },
    {
        "o":
      {                                   
        "text": "c",       
        "num": 44                             
      }
    }
]

I then execute the following query in PREVIEW mode:

SELECT CONCAT("coll"."o"."text", 'myString')
FROM mysource.mydb.coll

This runs fine – it returns all three “text” attribute values plus “myString”.
When I switch to “RUN” or execute via JDBC, I get:

Error: SCHEMA_CHANGE ERROR: New schema found and recorded. Please reattempt the query. Multiple attempts may be necessary to fully learn the schema.

Original Schema schema(o::struct<text::varchar>)
New Schema schema(o::struct<text::varchar, num::int32>)
SqlOperatorImpl MONGO_SUB_SCAN
Location 0:0:2
SqlOperatorImpl MONGO_SUB_SCAN
Location 0:0:2
Fragment 0:0

[Error Id: a37fca37-5532-462c-8f5e-d65d6e00a67c on MYMACHINE:31010]

  (org.apache.arrow.vector.util.SchemaChangeRuntimeException) Schema change error
    com.dremio.common.exceptions.UserException.schemaChangeError():88
    com.dremio.sabot.op.scan.ScanOperator.checkAndLearnSchema():261
    com.dremio.sabot.op.scan.ScanOperator.setupReader():178
    com.dremio.sabot.op.scan.ScanOperator.setup():163
    com.dremio.sabot.driver.SmartOp$SmartProducer.setup():560
    com.dremio.sabot.driver.Pipe$SetupVisitor.visitProducer():79
    com.dremio.sabot.driver.Pipe$SetupVisitor.visitProducer():63
    com.dremio.sabot.driver.SmartOp$SmartProducer.accept():530
    com.dremio.sabot.driver.StraightPipe.setup():102
    com.dremio.sabot.driver.StraightPipe.setup():102
    com.dremio.sabot.driver.Pipeline.setup():58
    com.dremio.sabot.exec.fragment.FragmentExecutor.setupExecution():347
    com.dremio.sabot.exec.fragment.FragmentExecutor.run():237
    com.dremio.sabot.exec.fragment.FragmentExecutor.access$800():88
    com.dremio.sabot.exec.fragment.FragmentExecutor$AsyncTaskImpl.run():594
    com.dremio.sabot.task.AsyncTaskWrapper.run():103
    com.dremio.sabot.task.slicing.SlicingThread.run():110

SQLState:  null
ErrorCode: 0

When I try exactly the same with a JSON source (i.e., same data, but no MongoDB), I get no error.

Query Profile:
query_profile_concat_bug.zip (41.8 KB)

Sample JSON file to import into MongoDB:
concat.zip (262 Bytes)

Thanks, Tim

2

Has anyone experienced something similar? Are there any workarounds? My team is working on a PoC and this issue is kind of a showstopper to us.

Just for completeness: I simplified the query to make it easy to reproduce. We’re actually trying to do some more complex operations involving parsing timestamp, etc.

Thanks, Tim

3

Hi @tid

Thanks for uploading the profiles. Let me look into this and get back to you

Thanks
@balaji.ramaswamy

4

Hi Tim,

Looks like we are running into a minor bug here. We have started to look at this via an internal Engineering ticket and will keep you posted on the progress

Thanks
@balaji.ramaswamy

5

Thanks, @balaji.ramaswamy!

Glad to hear that the issue is reproducible and under investigation!

Best, Tim

6

Hi @tid

Are you using the RPM or the TAR version?

Thanks
@balaji.ramaswamy

7

Hi,
We’re running on CentOS 7.5, so RPM.

Thanks, Tim