On 4/26/21 7:58 AM, Peter J. Holzer wrote:
On 2021-04-26 07:45:26 -0500, Ron wrote:
On 4/26/21 7:32 AM, Peter J. Holzer wrote:
On 2021-04-26 06:49:18 -0500, Ron wrote:
The destination is an (RDS) Postgresql 12.5 with encoding UTF8, and is being
loaded through COPY commands generated by ora2pg.
The source table has a BLOB column (I think they are scanned images) which
I'm loading into a Postgresql bytea column.
Seven times out of about 60M rows, I get this error:
Psql:909242: ERROR: invalid byte sequence for encoding "UTF8": 0xed 0xaf 0xbf
Decoding UTF8 doesn't make sense for a bytea column. How does that data
look like in the file generated by ora2pg?
I thought it was weird, too, but COPY has to read text, no?
Yes, but data for a bytea column would normally be encoded in hex or
something like that ...
COPY mv_response_attachment_old (response_attachement_id,binary_data,employer_response_id,attachment_id_code,file_type,attachment_desc,attachment_size,file_name,partition_date,prior_incident_id,part_date)
FROM STDIN;
1583201 \\x255044462d312e330d25e2e3cfd30d0a31362030206f...
... Yes, like this. There are only hex digits (plus \ and x) in the
column, nothing which would require decoding UTF-8.
My guess is that the error is actually in the data for another column.
I'd try to identify the broken records and check whether they contain
some other strange content.
That's a good idea. Thanks.
--
Angular momentum makes the world go 'round.
