Deployment Data Persistence
We need to take a hard look at the requirements of data storage, and how it needs to be implemented on a per-component basis.
- What services store data?
- What services share data stores?
- How do we share data stores? Persistent Disks, Object Storage, and ... ?
- What do we do when a PD is a Single Point of Failure?
- What services support HA?
- How is that HA implemented in relation to the data store(s)?
- How do we balance
git
data storage?- How do we balance storage so that the load is spread out?
- How do we create redundancy so a single pod or disk doesn't take out half a deployment?