Support multiple clusters to allow "paid tier" and "build osio on osio"

[maxandersen]

Multiple cluster support on osio are needed to allow:

a) for pay users who runs in another OSD
b) deploy of osio itself to another OSD
c) to scale beyond the initial user count there will be different clusters

open question are if these will be supporting Single-signon or we'll need to treat them as separate instances like i.e. github is separate from openshift.io access. Users wanting to target both openshift online and a different OSD would most definitely need that support.

[jstrachan]

@maxandersen for a) you mean a different OSO cluster right?
what does b) mean? If we're talking 100% on premise thats fabric8 right?
for c) you mean just for the users builds and runtime environments right? i.e. using a customers own OSD for their tenant apps and the Dev / Staging / Runtime environments while reusing everything else via the openshift.io SaaS?

[jstrachan]

details on changes required for multiple tenant clusters here: fabric8-services/fabric8-tenant#70

[alexeykazakov]

@jstrachan I think by c) @maxandersen means the case when a user's account may be provisioned to different OSO-starter clusters when for example the full capacity of the current (the only) east-2 cluster is used. Or the user is provisioned to a cluster geographically closer to the user.

[jstrachan]

so a) and c) are about different tentant clusters for tenant apps (jenkins/che), builds, pipelines and environments. Not the SaaS stuff.

I still don't grok b) though. You mean for hosting the shared SaaS services across multiple OSDs in multiple data centres? For things like fabric8-ui and Forge thats fine; for KeyCloak / WIT its much harder. Then fabric8-init-tenant we may need a pool of those per tenant cluster maybe?

[maxandersen]

b) is that ultimately we build osio components on osio and then need to deploy it back on to itself, rather than into openshift online.