[prebuild] prebuilds are marked as error even though they are in ready state

[sagor999]

Bug description

Here is a loom video showing an issue:
https://www.loom.com/share/55988106aa8a4fc39544a53a0611daaa

Here is a description from: #9710 (comment)

while trying to repro this case, I noticed an odd behaviour that I think is due to a fix for 
https://github.com/gitpod-io/gitpod/issues/9676 :
when prebuild has finished, and snapshot is being taken, prebuild will show up in dashboard in ERROR state. 
After snapshot was taken, it will get updated into READY state. But when I go into list of prebuilds,
it still shows up there as in ERROR state.

Steps to reproduce

follow the steps in loom video

Workspace affected

No response

Expected behavior

No response

Example repository

No response

Anything else?

No response

[geropl]

This is clearly a "misunderstanding" on the API level, about what intermediate conditions a workspace can be in. After #10053 got merged this should be clear.

When fixing this, we should extend our tests to not only cover singular status -> prebuild mappings, but also chains, and enforce constraints like "never go to error and back to available" etc. 👍

[laushinka]

Hi @sagor999, would you mind checking if this still happening? We had a bug in the fix for #9676 and the fix was just deployed this morning. I checked in the Admin dashboard for your prebuild-experiment project and the list of prebuilds show as ready.

[jldec]

@laushinka it looks like there may be other instances of this after the deploy this morning.
see slack (internal)

[sagor999]

@laushinka just tested and I am still seeing same behaviour. Prebuild is marked as failed after it is finished, and then once snapshot is created, it is updated to "ready".
When I go into prebuilds for the project though, I still see prebuild is marked as failed.

[DaniAkash]

@laushinka this is still happening in my workspace as well

[jldec]

NOTE:
With this issue, even though prebuild status is being incorrectly reported, prebuilds are succeeding, and workspaces should start properly. We are looking for a fix for the incorrect status.

[laushinka]

After the above PR is merged, we will also need to clean up the database.
In the prebuiltWorkspace DB, we should set the error field as empty when:
state is available and snapshot exists.

[AlexTugarev]

For completeness, we've chosen a simpler solution for now as discussions towards #10054 (comment) would not be quickly done.

When fixing this, we should extend our tests to not only cover singular status -> prebuild mappings, but also chains, and enforce constraints like "never go to error and back to available" etc. 👍

That again would require to re-introduce computation of state updates based on existing state and incoming events.