For discussion: some breaking changes

[gilesknap]

The Last of the Breaking Changes for epics-containers

Intro

A few things that require breaking changes have been mounting up. Here I'm trying to collect everything I can think of it so we can have one last go at making the framework tidy, before it gets fully out into the wild and is no longer mutable.

Goals

• Make helm charts look like standard patterns and easy to understand
• Have all helm charts able to share the common fields from helm create (tolerations, labels etc.)
• standardise on the method of making and publishing helm values schema
• Make it easy and tidy to make a fastCS helm chart that works as an IOC or non IOC inside an epics-containers services repo or outside of one
• make the PVC ArgoCD App have a prefix in its name so it can be deployed to the Accelerator with multiple services repos
• make the ec tool more generally useful and not IOC specific
• make tracking main in ArgoCD a possibility
• Add a way to publish techUI assets to IOC instances (also use this mechanism for extra soft support)

Things to Change

• Look hard at the config folder hashing problem - do we need a breaking change to make that work?
• move epics-opis and epics-pvcs:
  • implement backup for autosave PVCS and execute for those IOCs that need it
  • Rename epics-pvcs ArgoCD application to ixx-pvcs (required for accelerator)
  • Rename epics-opis ArgoCD application and service to ixx-opis (required for accelerator)
  • Change all PVCs so that their PVs are retained when the PVC is deleted
  • restore backed up autosave
• remove all of the the generated variables at the top of the ioc-instance helm chart such as $opisClaim $location etc
  • instead explcitly add them to the shared values.yaml
  • if they are missing then don't declare a PVC and don't mount
  • or some better pattern for discussion
• rename iocEnv and globalEnv values to env and .global.env and make the former override the latter
• structure ec-helm-charts/Charts like fastcs/Charts and add the helm schema precommit
• Look hard at the config folder hashing problem - do we need a breaking change to make that work?
• Also look at the techUI publishing problem.
• remove ec-services from ixx-deployment/apps/values.yaml and only have services
  • then: have ec report it cannot stop services that dont support enabled
  • use a label on the pod to declare it supports enabled
• UPDATE DOCUMENTATION to reflect the above (internal and external docs)

Required Improvements to ec

• make argocd login happen on any command when cached creds are out of date
  • and cope with running over ssh from home (how?)
• remove the confusing environment.sh or at least make it work in DLS (from services and deployment repos?)
• make ec work for IOCs and non IOCs
• add exec and attach to ec (requires that the argocd backend can also call kubectl)

[gilesknap]

@GDYendell @coretl for discussion on Monday.

I might need to go on a retreat to get these done!