Creating a large dataset by crawling some public website

I need a large dataset in JSON format, to upload in a MongoDB database. The contents can be anything, but should be meaningful. I need between 500 MB and 5 TB of data to be generated. The data will be used for some training demonstrations. I want someone to write a program that crawls some website for publicly available data (such as books and reviews from some e-commerce site; news articles from some news sites; hotels and reviews from some travel site; restaurants and reviews from some food aggregator site; articles from wikipedia, etc). I don't need you to send me the data. I need you to write a program I can run at my end to download the data. But the program must store it in a JSON format that can be directly imported into MongoDB. The structure could be flat JSON documents, or documents that contain embedded documents. Individual documents may be anywhere in the range from 100 bytes to 100 KB. No individual document should be bigger than 100 KB in size. We'll have to discuss together to decide the site from which the data is to be downloaded. There should be no violation of any data access policies of the site. This is very important for me; I don't want us to break any law. I will need an assurance from you on this, and a link to the data access policies of the site, if available. Once we agree on the site to download the data from, you will write the program, test it at your end, send me some sample data, and once approved, send me the program for me to run at my end. If I run into any difficulties while running the program I would require you to support me. The program should allow me to choose the approximate data size (such as 500 MB) after which it will stop crawling any further to download the data.