I'm trying to script a Personalize CreateDatasetImportJob. The (redacted) JSON payload looks like so:
{
"jobName": "some-job-name",
"datasetArn": "my-dataset-arn",
"dataSource": {
"dataLocation": "s3://bucketname/path/to/my/data/date=20230801/"
},
"roleArn": "my-role-arn",
"importMode": "FULL"
}
When I run this job via the console, with that same S3 URL (prefix), the job runs just fine. When I execute via the CLI using that payload, however, I get this error:
No valid input at s3://bucketname/path/to/my/data/date%3D20230801/
Personalize seems to be URL-encoding the "=" in my (perfectly valid) S3 URL path. Any ideas how to prevent this? Changing the path structure of my data is not an option ... these are carefully controlled data lake paths -- that work with all other AWS tooling so far ... only Personalize seems to be giving me grief, and only programatically, as the console is working but doesn't scale for thousands of jobs :-)
