Spectrum Scan Error - Unable to create parquet column scanner

Question

Hi, I am trying to load a simple 13 columns table from s3 parquet to Redshift table. Not using spectrum or external tables etc, but looks like reading parquet is spectrum in the background anyway. I am getting same error on other tables too.
I am keep getting this error below, re-created cluster, rebooted as well but error is still there. Could someone tell me where can i find the root cause of this error?  I am sure iam role is picked up, file is there,  column count match etc as I fixed all those errors and got this last error that I can not understand the cause of it.

```
ERROR: Spectrum Scan Error Detail: ----------------------------------------------- 
error: Spectrum Scan Error 
code: 15001 
context: Unable to create parquet column scanner 
query: 525 
location: dory_util.cpp:1167 process: worker_thread [pid=27989] ----------------------------------------------- 
[ErrorId: 1-6204d783-12b35e0914588ba01e3c14fe]

```
copy command I use, which is simple
```
 COPY my_schema_name.mdvm					
FROM 's3://bucket/data/table/files/'							
IAM_ROLE 'arn:aws:iam::xxxxx:role/service-role/AmazonRedshift-xxxxxxxxx'							
FORMAT AS PARQUET;
```

Answer

It would be best to open a support ticket so the files can be investigated. I have seen instances where decimal value was encoded differently than Redshift expectation, encoded as `binary (STRING)` instead of `fixed_len_byte_array(5) (DECIMAL(10,4))`. If this is the case then you can convert the string values to decimal with following Pythin code
```
import pyarrow as pa
new_col = pa.chunked_array(pa.Array.from_pandas(table.column('col1').to_pandas().astype(float)).cast(pa.decimal128(10, 4)))
table = table.set_column(position, 'col1', new_col)
```
Hope this helps!

Spectrum Scan Error - Unable to create parquet column scanner

관련 콘텐츠