Bulk restore-request of objects in S3.

glacier-bulk-restore.sh

#!/bin/bash

BUCKET_NAME="my-bucket"
KEY_PREFIX="my-key-prefix"
DAYS = "10"

aws s3 ls --recursive "s3://$BUCKET_NAME/$KEY_PREFIX" |\
  awk '{$1=$2=$3=""; print $0}' |\
  sed 's/^[[:space:]]*//g' |\
  tr '\n' '\0' |\
  xargs -t -0 -I {} aws s3api restore-object --restore-request Days=$DAYS --bucket "$BUCKET_NAME" --key "{}"

Iterating over columns with awk is problematic for object keys with reserved characters. A resilient approach is blanking the date, time, and size columns, printing the remainder, then removing any leading whitespace left over. Following that, xargs is not robust against quotes (producing “xargs: unterminated quote” errors). Convert newlines to null, then use -0 to indicate null-delimited lines.

See also:

• Using awk to print all columns from the nth to the last
• Howto Make Find Xargs Grep Robust to Spaces in Filenames on a Mac