Your design looks fine I think. It depends on how you are going to retrieve them. In your example you have a blog system so I assumed you need to retrieve your blog alone with all its comments (and sub-comments as well), then you could have blog id as the partition key and retrieve all comments in one query and it also ensure all comments under the same blog are stored in the same data cluster in azure data center with max performance.
If you need more performance I'd like suggest you to store author name as well, to reduce the join operation in your application layer.