Re:Why Uber Engineering Switched from Postgres to MySQL
by sam_pointer on Wednesday July 27, 2016 @03:56AM
(#12171221)
Unfortunately not. That company and codebase is very much dead, and I don't own any of it. Even if I did it would be a bunch of context-less shell (in the first iteration) and then a bunch of context-less Chef (in the second platform iteration).
Things I do remember:
- We used ephemeral volumes for all data stores. This was pre-provisioned IOPs and EBS was flaky as heck back then. I can't remember the disk layout, although we did experiment a great deal.
- We took great pains to ensure there was enough space to make the snapshot (IIRC is was a telemetry/monitoring item)
Things I do remember:
- We used ephemeral volumes for all data stores. This was pre-provisioned IOPs and EBS was flaky as heck back then. I can't remember the disk layout, although we did experiment a great deal.
- We took great pains to ensure there was enough space to make the snapshot (IIRC is was a telemetry/monitoring item)
- The pipe scripts were essentially "netcat".
The best I can offer is this talk: http://vimeo.com/57861199