From a VPS to AWS in two steps: migrate first, modernize after
- Sector
- Pharmaceutical
- Duration
- 2 days to rehost · 1 week to modernize
- Role
- Proposal, design and delivery
01Where it started
The application, the MySQL database and Redis for sessions and cache shared a VPS, and users' files sat on its disk. One failure took everything, and it had to move out with little time.
02What could not move
- Time was short: the migration could not wait for the modernization.
- A container can be replaced at any moment: it cannot keep anything on its disk.
- All of the infrastructure is written as code.
03The decision
Two steps instead of one. First, rehost: the same application on an AWS instance, unchanged, to meet the date. Then, with the clock off, replatform: containers, managed MySQL and Redis, and files on object storage. Everything holding state left the server, which is what lets a container be replaced without losing anything. Doing both at once was a single cutover, with two large changes together and no way to tell which one broke what.
04The trade-off
- What it cost
- It was migrated twice: two cutovers and two data checks. And for a while, AWS ran the same single machine the VPS had, with the same risks.
- Why it was worth it
- The first step took the date out of the way and the second was done calmly. Separating the migration from the change of architecture tells you, if something fails, which of the two it was.
05What came of it
The application no longer depends on one machine: containers are replaced without losing anything, because the database, the sessions and the files live outside them. Deploying is publishing a new image, and the infrastructure is written as code.
Something similar in front of you? Tell me what you are dealing with and I will reply, almost always within 24 hours.
Get in Touch