Presented by:

Headshot

Morgan Howes

Procore Technologies

I am a Postgres DBA that gets to have fun solving the challenges that a hyper-growth unicorn SaaS company provide.

When there is a break from work, I ski, scuba, and generally try to be outside.

No video of the event yet, sorry!

Join me in recounting the trials and tribulations of moving a 14 terabyte monolith database from a bare metal, on premise solution to Amazon Web Services (Aurora/RDS). Specifically, the focus is on how to move the data with high availability, minimal downtime, and still leave the option to go back to the old hardware.

This combination of case study and tutorial goes into the technical details of how we worked through one issue after the other until finally making our way to Amazon. Included are looks into:

  • The hardware in both the old and new environments.
  • The challenges of replicating into RDS/Aurora clusters.
  • How to dump and restore our 25,000 tps database in under a week.
  • Coordination, timeline estimation, and testing while you still have a day job.
  • The many lessons we learned using Bucardo as a replication tool.
  • How to fold an Integer>Bigint migration into the database move.

Several coworkers and I sweat and cried over this process for months, all so you could have this easily digestible talk (ok mostly to help our company to continue growing). So learn how to move large databases into the future, or just laugh at our pain and know why you want to do this sooner than we did.

Date:
2019 September 19 15:00 PDT
Duration:
50 min
Room:
Winchester (2)
Conference:
Silicon Valley 2019
Language:
Track:
Ops and Administration
Difficulty:
Medium