Use Glacier to snapshot S3 like a point-in-time backup

I see some useful comments are already here but will throw in just one comment: I would frankly go with putting your objects into S3 and then simply create a very short "Transition to Glacier" policy, like a few hours or 1 day old. You can access Glacier directly through the API, but the reason I would go this route is that you will still see your objects listed within S3, but you'll see in the "Storage Class" column that after a few days everything will have been moved into Glacier. Then, when/if you need to retrieve something out of Glacier, you can make the retrieval request (called "initiate restore" in the new ACTIONS menu of S3's web GUI) it's all right there for you to see and work with.

I'm a huge fan of the CLI for much of my AWS work, but I think until there's a really robust and usable CLI tool for Glacier, this seems easier to work with.

You can use this:

mysqldump \
    --host=127.0.0.1 --port=1234 \
    --user='username' --password='password' \
    --single-transaction --routines --triggers --compress --compact \
    --databases '<database name>' | gzip | aws s3 cp - s3://'<bucket name>'/'<database name>'.sql.gz