Postgres Is the New Default

Postgres is the new default – How we transitioned our platform at Engine Yard and why you should too.

Given at Postgres Open 2012. Video is available here: http://www.youtube.com/watch?v=sFfJjfOk9NA

Transcript

1. Postgres is the new default How we transitioned our platform and why you should too
2. I work with Data!
3. Agenda • Where we were • Our relationship with Postgres • Challenges • Our future together
4. We live in The Cloud!
5. Engine Yard • Languages • Amazing Support • Team DBAs & me • 2 Products *
6. > 1 Petabyte
7. No multi-tenancy.
8. ...A year (or so) ago
9. Engine Yard Cloud • Only MySQL 5.0 • 8.3 in perpetual Alpha • No Slaves, No Docs • Increasing interest • Stack Expansion - DIRTT
10. Our Relationship with Postgres
11. Infatuation Stage 1
12. Why love Postgres? • Flexible Replication: streaming & hot standby • Features & Extensions • Versatile Data Types • FDWs
13. Why love Postgres? • All of the Indexes! • B-tree, GiST, GIN, SP-Gist* • Procedural Languages (PL) • Full-Text Search
14. Why love Postgres? • Feature parity with Managed • Outstanding community & commercial support • Let’s make it our default!
15. Understanding Stage 2
16. Re-Tool • MySQL Shop • DBA Training: DBAs & Engineers • OSS to help us gain velocity
17. Customer Education • Customer Facing Docs focused on MySQL • Technical Writers FTW!
18. Disturbances Stage 3
19. Roadmap issues • Shiny new process for DB releases • Alpha, Beta, GA • Surprise we got it WRONG!
20. Multiple Architectures • EBS Volumes & 32-bit instance sizes • Dump & Restore only way to vertically scale
21. Bad Assumptions • Product codebase tied to old assumptions • MySQL everywhere • Refactor!
22. Moulding Stage 5
23. Company-Wide Effort
24. Installation • Engine Yard Cloud • Chef > EBS • Engine Yard Managed • Puppet > Storage
25. Configuration • $PGDATA on EBS • We don’t RAID • Snapshots to S3 • pg_xlog same volume
26. Replication • Archiving on db_master • hot_standby slaves • All slaves <> AZs • Smaller slaves
27. Monitoring • Bucardo’s check_postgres • Collectd & Nagios • Process health, replication, checkpoints - MORE
28. Backups • Internal tool wraps around pg_dump • Snapshots & dumps to S3 • Only backup from slave (if present)
29. Allow Extensions • Dedicated instances • Customizations OK! • Ship recipes for ~30 extensions • Curated & tested
30. Architectures • New instance types: 64-bit small, mediums • Standardized instances architecture 64-bit
31. Happiness & Trust Stage 9
32. Future Directions • 9.2 Support • AWS Optimizations • Provisioned IOPS • EBS optimized instances • SSD-backed instances
33. Future Directions • Painless upgrades • Better extensions • Ugh to cookbooks! • HA/DR
34. Future Directions • Customer education still a problem • Low adoption of slaves • ~146 year ago • ~165 now
35. Give us a try!
36. Questions? @RandomMood