This patch addresses Nate's feedback that it's difficult to know what
platform is being operated on.
Previously it wasn't clear where the platform id used for push and pull
comes from. The source of truth is the platform.metadata.json file
created when the platform is first generated using `holos generate
platform k3d`.
This patch removes the platformId field from the platform.config.json
file, renames the platform.config.json file to platform.model.json and
renames the internal symbols to match the domain language of "Platform
Model" instead of the less clear "config"
This patch also changes the API between holos and CUE to use the proto
json imported from the proto file instead of generated from the go code
generated from the proto file. The purpose is to ensure protojson
encoding is used end to end.
Default log handler:
The patch also changes the default log output to print only the message
to stderr. This addresses similar feedback from both Gary and Nate that
the output is skipped over because it feels like internal debug logs.
We still want 100% of output to go through the logger so we can ensure
each line can be made into valid json. Info messages however are meant
for the user and all other attributes can be stripped off by default.
If additional source location is necessary, enable the text or json
output format.
Protobuf JSON:
This patch modifies the API contract between holos and CUE to ensure
data is exchanged exclusively using protojson. This is necessary
because protobuf has a canonical json format which is not compatible
with the go json package struct tags. When Holos handles a protobuf
message, it must marshal and unmarshal it using the protojson package.
Similarly, when importing protobuf messages into CUE, we must use `cue
import` instead of `cue go get` so that the canonical format is used
instead of the invalid go json struct tags.
Finally, when a Go struct like v1alpha1.Form is used to represent data
defined in cue which contains a nested protobuf message, Holos should
use a cue.Value to lookup the nested path, marshal it into json bytes,
then unmarshal it again using protojson.
Previously I developed holos server in the dev-holos namespace of a
remote cluster. This patch updates the Tilt configs to develop locally
against k3d quickly and easily.
The database is a CNPG database which replaces PGO. This is simpler and
ligher weight, one container in one pod. CNPG has no repo host like PGO
has.
Previously, the Tiltfile was hard-wired to Jeff's development
environment on the k2 cluster on-prem. This doesn't work for other
contributors.
This patch fixes the problem by re-using the [Try Holos Locally][1]
documentation to create a local development enironment. This has a
number of benefits. The evaluation documentation will be kept up to
date because it doubles as our development environment. Developing
locally is preferrable to developing in a remote cluster. Hostnames and
URL's can be constant, e.g. https://app.holos.localhost/ for local dev
and https://app.holos.run/ for production. We don't need to push to a
remote container registry, k3d has a local registry built in that works
with Tilt.
The only difference presently between evaluation and development when
following the local/k3d doc is the addition of a local registry.
With this patch holos starts up and is accessible at
https://app.holos.localhost/
[1]: https://holos.run/docs/tutorial/local/k3d/
Add Tilt back from holos server
Note with this patch the ec-creds.yaml file needs to be applied to the
provisioner and an external secret used to sync the image pull creds.
With this patch the dev instance is accessible behind the auth proxy.
pgAdmin also works from the Tilt UI.
https://jeff.holos.dev.k2.ois.run/app/start
