I’ve been deploying KNIME on AWS EC2 instances for image analysis. If I write directly to S3 (using table writer node in the workflow), it is very slow but the file is properly recognized. If I write the file to the EC2 instance SSD and than manually transfer the written table to S3, the table is not recognized when I try to open it in Table Reader. I subsequently get a “no entry data.bin in file” error. Also, the files are no longer characterized as Zip archives in Ubuntu but marked as “unknown”.