How does the unicorn work

Presentation shwon on 19.02.2015 at Łódź Ruby user Group

Transcript

1. How does the Unicorn work? Michał Gutowski Binar::Apps m.gutowski@binarapps.com / ja@null.yt
2. Agenda • Basic info • Rack • Connection handling • Forking & load balancing • Parsing HTTP requests • Restarting
3. Basic info • Commonly used web server for RoR applications • Rack based • uses traditional Unix prefork mode • and Unix signals
4. Rack • Interface between Ruby and Ruby frameworks • Wraps HTTP requests and responses • You can exchange rack based severs without changing application
5. Connection handling • Master process creates listening socket • When new connection arrives child process and new socket are created • Operating system takes care of load balancing
6. Parsing HTTP requests • Taken from Mongrel • Uses Ragel state machine
8. Restarting • Possibility to hot swap • In case of failure possibility to roll back 1. USR2 to master process 2. WHINCH to old master 3. QUIT to old master * in case of trouble HUP 
 to reload old master
 and QUIT to new master
9. References • https://github.com/blog/517-unicorn • http://tomayko.com/writings/unicorn-is-unix • http://unicorn.bogomips.org/
10. ujeb.se/JMVA