S3: top 20 biggest objects in a bucket

Bill surprises from S3 are almost always a few giant objects. This finds them.

$ aws s3api list-objects-v2 --bucket my-bucket --max-items 5000 \
  | jq -r '.Contents | sort_by(-.Size) | .[0:20] | .[] |
             [.Size, .LastModified[0:10], .Key] | @tsv' \
  | awk -F'\t' '{ printf "%10.2f MB  %s  %s\n", $1/1024/1024, $2, $3 }'

1. [01] awsaws s3api list-objects-v2 --bucket my-bucket --max-items 5000
   list-objects-v2 paginates at 1000 keys per page; --max-items caps the total. For huge buckets, switch to S3 Inventory (a daily manifest delivered to another bucket).
2. [02] jqjq -r '.Contents | sort_by(-.Size) | .[0:20]'
   Sort descending by Size, take top 20. Negative sort is jq's idiom for descending.
3. [03] awkawk -F'\t' '{ printf "%10.2f MB %s %s\n", $1/1024/1024, $2, $3 }'
   Convert bytes to MB with 2-decimal precision and right-align. The one awk trick worth knowing for size formatting.

   4218.42 MB  2026-04-23  exports/full-backup-2026-04-23.tar.gz
   2841.07 MB  2026-05-09  measurements/raw/run-2026-05-09T14.parquet
   1936.55 MB  2026-05-08  measurements/raw/run-2026-05-08T22.parquet
    902.31 MB  2026-05-01  exports/monthly-2026-04.tar.gz

• For buckets with >5000 objects, use S3 Inventory — list-objects-v2 gets slow and expensive on huge buckets.
• Pipe the Key column through `claude -p "Suggest a lifecycle policy that would have prevented these from accumulating"` for a quick cleanup plan.