Quick general question about GDCN... when designin...
# gooddata-cn
Quick general question about GDCN... when designing a developer workflow... would one recommend having the developer's local machine run k8s and GDCN.CE in a docker image in that for local development... or, try to load the whole GDCN k8s install on the developer's machine. I'd like for the developer's experience to be as much like production as possible, but loading the GDCN k8s locally seems excessive. For the rest of the system... I think I need to run it in local k8s.
There are multiple ways how to achieve convenient workflow. First, a few questions should be answered: What is the target persona of this workflow? What needs to be developed - new metrics, reports, dashboards? Or some more complex automation processes, upgrade steps etc? Does the environment need to run all the time? Does it need to be accessed concurrently by multiple users? If you want to prepare metrics/reports/dashboards, running on GDCN.CE (AIO) may be sufficient if you have connected datasource with the same schema (and possibly the same data). You may export workspace and load it to production. Having k8s on localhost doesn't bring too much benefits compared to AIO except the following use-cases: • multiple organizations • PDF exports Both locally running AIO and k8s suffer the same issue - this deployment is not remotely accessible so you can't collaborate with other people and remote integrations will now work as well. Having fully-fledged k8s development deployment in public/private cloud resolves all these issues (remote access, pdf exports, multiple organizations, upgrade tests) at the expense of extra hardware costs. This issue can be reduced by lowering replica count or by suspending computes when it is not needed. This development environment may share the same k8s cluster as production cluster (it can share the pulsar as well). If the extra costs are still unacceptable, you may create only "development organizations" within the production environment. This solution will not cost you a penny but there are other tradeoffs: development can possibly affect production performance, you can't use it for upgrade tests.