On a fresh one-click deploy, the preflight is always red, leading to a poor UX

[maelvls]

There is still a discussion we are having with @wallrj about the poor UX regarding the failing preflight deployment: when a user clicks "Deploy", he should be expecting everything green. But currently, the preflight deployment stays failing until the user proceeds with step 2. What I did to remediate this is to explain to the user in step 1 that the failing preflight deployment is expected and that they should go on with installing the preflight agent to get it fixed.

[maelvls]

I received this email (jetstack internal link) where Dinesh, who is in charge of reviewing our app before it can be published, asks me to fix our app because he sees that the preflight deployment fails.

I deployed your listing on to one of our test clusters and noticed that the prefligt deployment is failing with the following error

configmap "agent-config" not found

Can you please update the ConfigMap name to "ubbagent-config" instead of "agent-config" in the preflight deployment and resubmit the draft for a review.

In my previous email, I had preemptively told him that the preflight deployment was expected to fail and that it was part of our "normal" flow and that it is documented in the user guide:

I just re-submitted the application with the new release 1.1.0-gcm.6. This fixes the notes and description on the Application view. Note that the preflight deployment is expected to be failing on a fresh click-to-deploy or CLI installation.

[maelvls]

Plan: have the preflight deployment scaled to 0 by default. I'll change the user documentation: when the user wants to enable preflight, they will have to type one more command (in README.md):

 sed '/namespace:/d' agent-config.yaml | kubectl -n $NAMESPACE apply -f-
 kubectl -n $NAMESPACE rollout restart $(kubectl -n $NAMESPACE get deploy -oname | grep preflight)
+kubectl -n $NAMESPACE scale --replicas=1 $(kubectl -n $NAMESPACE get deploy -oname | grep preflight)

Timeline: I will have a new release 1.1.0-gcm.8 by this evening and re-submit to Google.

FYI @mattbates @wallrj @charlieegan3

[charlieegan3]

I think this works for now. Ideally we have a tighter integration here in future.

[charlieegan3]

@mattbates this seems like a clear priority for the next jscp sprint.