vfulco / WeiboSpider

Sina Weibo Scraper by Yong HU and Mingyang LI. 2,000 posts per second.

WeiboSpider

This is a sina weibo spider built by nghuyong largely tailored to run on WWBP's servers by Mingyang Li.

A detailed explanation, written by nghuyong, can be found at 微博爬虫总结：构建单机千万级别的微博爬虫系统.

Description of data structure can be found at 数据字段说明与示例.

Other Branches

The original repo by nghuyong has 3 branches:

Branch	Structure	Posts per Day
simple	single account	100,000
master	account pool	1,000,000
senior	distributed pool	10,000,000

Usage

1. Clone thre repo. Install dependencies.

   git clone [e-mail]:nghuyong/WeiboSpider.git
   cd WeiboSpider
   pip install -r requirements.txt

2. Install phantomjs, mongodb, and redis. Start the latter two.
3. Write down the usernames and passwords of some Sina Weibo accounts in sina/account_build/account.txt. Follow the format indicated in account_sample.txt.
4. Populate the account pool by running python sina/account_build/login.py.
5. Populate URLs to start scraping with by issuing python sina/redis_init.py.
6. Run scraper by running scrapy crawl weibo_spider.

Data Storage

Posts, user profiles, and user relationships (and comments optionally) are stored in the MongoDB.

Performance

With the default setting, 16GB memory, 8-core CPU, Ubuntu, and 36 processes, we are hitting an average of 2,000 posts per second.