Rails Usergroup Hamburg: Heroku
Download as KEY, PDF0 likes700 views
Heroku is a cloud platform that allows companies to host their Rails applications without owning the physical infrastructure. Some key advantages are variable costs that scale with usage and high flexibility. Deployment can be challenging due to limitations on configurations and reliance on a third party. The Heroku stack provides automatic scaling through dynos and add-ons, and uses Git-based deployment where code is pushed to build application slugs. While not suitable for all large applications, Heroku is recommended for new simple projects due to its low cost and ease of use.
Rails Usergroup Hamburg: Heroku
- 1. Heroku Rails Hosting in the Cloud
- 2. What is cloud computing?
- 3. Cloud Computing • Companies don‘t own the physical infrastructure • Computing time and storage is rented from a third party
- 4. Advantages • Variable costs instead of capital expenditure • Costs lining up with real usage • Real usage lining up with sales volume • Higher flexibility
- 5. Disadvantages • You have to trust a third party • Limitations for exotic configurations
- 6. The Cloud rocks Deployment sucks
- 7. co sts agility low scala bility The Cloud rocks abso rb p ur ity eaks s ec reliability
- 8. Set up new Deal with instances outages Deployment sucks Persist Balancing storag ant Load e
- 11. Customers contract Dyno, SQL-Storage, Caching,… management ECS, S3, EBS, SimpleDB, …
- 12. Customers contract contract manages ECS, S3, EBS, SimpleDB, AppServer, SQL, etc. …
- 13. The Heroku stack
- 15. OS and Ruby • default: Debian 4 + Ruby 1.8.6 • optional: • Debian 5 + REE 1.8.7 • Debian 5 + Ruby 1.9.1
- 16. Dynos App servers, background processes
- 17. Database Shared or dedicated, based on PostgreSQL
- 18. Add-Ons • Amazon RDS • Websolr • Bundles (Backups) • Memcache • Cron • AMQP
- 19. Integration • Custom domains • New Relic • DNS servers • Sendgrid • Deploy hooks • SSL certificates • Exceptional • ...
- 20. Read Only FS • Caching → Header + Reverse Proxy • Uploads → S3 • SQLite → PostgreSQL • Ferret indices → use SOLR service
- 21. Interaction with heroku
- 22. Git based workflow • gem install heroku • heroku create • git push heroku master • heroku rake db:migrate
- 23. git push → build slug } • checkout master branch • download, build and install gems slug • delete .git, log/*, tmp/* and everything in .slugignore
- 24. Gem management • .gems manifest file • bundler (experimental)
- 25. Database management • Uses values from database.yml • Export to heroku → db:push • Import from heroku → db:pull
- 26. Other features • heroku • console (interactive/single commands) • rake
- 27. Conclusion
- 28. Cons • Oversimplification • Predefined sets of OS and Ruby version • No ssh access • No choice of physical location • Only predefined tools available • For larger setups: pretty expensive • Porting apps may be hard
- 29. Pros • Very simple start • Low costs in the beginning / for small apps • Offers simple management of complex apps, if developed with the heroku stack in mind
- 30. I do recommend Heroku for any new projects that were built with heroku restrictions in mind, because it can‘t get any simpler!
Editor's Notes
- #5: * no capital expenditure * low barriers to entry * low risk (cancel contract any time) * e.g. fire up test systems for load tests and pay by the hour * utility like electricity
- #6: * e.g. Amazon and possibly your management provider can read your data * XING uses SSDs to quickly serve small image files (not possible with Amazon)
- #8: * agility/scalability: fire up instances automatically * low costs: variable costs, low risk * security: experts take care of backups, it‘s their core business * reliability: replicate data between cloud providers, fire up instances on the other if one is down
- #10: The cloud does not make any sense without automation.
- #12: * no direct contact between customers and amazon * heroku could switch the hosting provider any time without anyone noticing
- #13: * customer has 2 contracts: one with scalarium, one with amazon * scalarium is tailored specifically to aws
- #15: * Reverse Proxy: Nginx * HTTP Cache: Varnish * Routing Mesh: Custom, written in Erlang * SQL: PostgreSQL * Memcache
- #18: * shared: multiple users on 1 powerful computing unit ** better for peaks ** impacts by other users * dedicated: 1 dedicated computing unit * database.yml will be replaced automatically on deployment
- #23: heroku create: git remote add heroku git@heroku.com:my_app.git http://my-url-35.heroku.com/
- #24: .slugignore format like .gitignore
- #25: * current bundler version on heroku: 0.9.11
- #29: * No Ruby 1.9.2 -> no Rails 3 beta3 * e.g. no Redis * every custom tool must be manually managed, out of the Heroku scope * technical term confusion, e.g. bundle and migration