Jason Root explained that Tamu is working on an implementation timeline for Folio - with a go-live goal of Summer 2020.
- Production-ready K8s/Rancher infrastructure needs to be in place by the Fall 2019.
- Decided on a stand-alone production K8s cluster for live Folio, with separate clusters for Folio testing, Folio training, other projects and so on.
- Using Rancher Server for control plane, RancherOS for the K8s nodes and VMware provisioner.
Anton Emelianov discussed benefits of CI/CD deployments using K8s and multi-tenants. Current CI/CD builds from merging on Master, with a single tenant, can create problems - as we've all seen.
Taras Spashchenko outlined the work he has been doing with module schema versioning, upgrading of schemas and data when a back-end module is upgraded and introduces db changes.
- Started with simple use-case, changing the structure of the jsonb.
- Modules should be in charge of validating and migrating the data between themselves.
- Anton Emelianov relayed that Hongwei Ji suggested for a module to connect to more than one db, so data migrations could happen from one db instance to another in the event of jsonb structure or schema changes.
- At present, a back-end storage module cannot connect to more that one db at a time. If this needs to change, a lot of work will have to be done from the development side. Perhaps there could be a single "migration" module that could do this for storage module schema upgrades?
- Currently, no way to validate the data from within Postgres once it has been migrated. Maybe there are Postgres extensions for this?
Ian Hardy updated us on the core-team's work with K8s integration with Jenkins CI/CD. There are several tickets in Jira that are being looked at.