extract_table() yields 403 error on GCS bucket with retention policy

[kansuke-at-trimble]

extract_table() yields 403 error on GCS bucket with retention policy

[kansuke-at-trimble]

Hi @Linchin, thanks for the help! It turns out that client.extract_table already created parquet files and the line extract_job.result() tried to do the exact same thing which caused the retention policy error. I deleted the extract_job.result() as it seems to be unnecessary and now the job runs correctly. Thanks again, and since it's solved on my end I am closing this issue

[Linchin]

Thank you @kansuke-at-trimble for raising the issue! From what I can see, it seems to be working as intended. I suppose 'bucket-a/train_dataset/000000000124.parquet' already exists, and a retention policy indicates that

objects in the bucket can only be deleted or replaced once their age is greater than the retention period[1].

So basically the file cannot be overwritten during the retention period. I guess a way to bypass it is to use a unique file name, maybe one with UUID or the timestamp when it's generated.

Also, the client doesn't do the extract job itself - it sends the command to the backend, and the backend does the work. That's why there's no code in the client library doing the extraction. Hope these answer your question, and please let us know if you have any further questions :)