[supervisor]: gitpod config changed analytics

[akosyakov]

Description

This PR adds analytics for .gitpod.yml changes. We are track when changes are persisted to a disk and debounce them for 5s. After last change we analyse changes to top-level elements and report an event for each. It allows us to gain insights in what users are changing on top-level, e.g ports or tasks, and avoid a lot of traffic at the same time. Later we can do more fine grained analysis to see which object properties, like ports.name.

I had to resolve couple issues which were discovered during review or were blocking testing:

• enable analytics env inside workspace in preview environments to verify with segment
• don't fire change events for syntactically broken .gitpod.yml files, thank you @andreafalzetti for noticing it
• adding git hook which keeps goland types up to date with gitpod schema, otherwise we will miss new properties, thank you @mustard-mh for brining forward that yaml parser only see properties which are defined in goland

Related Issue(s)

Fixes #6894

How to test

• Open staging untrusted.
• Start a workspace, open .gitpod.yml yet, wait for 5s.
• You should not yet see any gitpod_config_changed events.
• Now edit .gitpoy.yml, then you finish wait for 5s.
• Now you should see gitpod_config_changed events.
• Edit .gitpod.yml to add some syntax errors and semantic changes, then you finish wait for 5s.
• You should not yet see any gitpod_config_changed events.
• Edit .gitpod.yml to fix syntax errors, then you finish wait for 5s.
• Now you should see gitpod_config_changed events for semantic changes.

Check for regressions:

• Add new ports and check there are detected in the ports view.

Release Notes

NONE

Documentation

Werft options:

• /werft with-preview
• /werft analytics=segment|TEZnsG4QbLSxLfHfNieLYGF4cDwyFWoe

[werft-gitpod-dev-com]

started the job as gitpod-build-ak-identiy-config-changed.5 because the annotations in the pull request description changed
(with .werft/ from main)

[werft-gitpod-dev-com]

started the job as gitpod-build-ak-identiy-config-changed.7 because the annotations in the pull request description changed
(with .werft/ from main)

[akosyakov]

/werft run with-clean-slate-deployment=true

👍 started the job as gitpod-build-ak-identiy-config-changed.9
(with .werft/ from main)

[akosyakov]

This file auto-generated. I added git hook which regenerates it on commit, see .pre-commit-config.yaml

[akosyakov]

Refactored to extract a common properties to the definition and stabilize code generation.

[felladrin]

Working as advertised 🎉

I left just a comment about events triggered when the workspace has no .gitpod.yml and we run gp init, which is an edge case.

[akosyakov]

I left just a comment about events triggered when the workspace has no .gitpod.yml and we run gp init, which is an edge case.

let me verify first that it does not happen then a user creates an empty file as well, it does not look good

[akosyakov]

@felladrin I pushed the fix when https://werft.gitpod-dev.com/job/gitpod-build-ak-identiy-config-changed.38 is ready you can try again.

[andreafalzetti]

Great work @akosyakov 🙏

It works as expected when the .gitpod.yml is invalid. I've noticed that when creating with gp init or deleting the whole file it does detect changes for each field but I'd say it's kind of expected and I don't consider it a bug.

[akosyakov]

@loujaybee @andreafalzetti Do you think it would be useful to add an attribute like operation to distinguish what happened, i.e. added, updated, removed? I looked right now at analytics it can be misleading, i.e. you can see 2 modifications to tasks, but final maybe was removing so actually tasks are not used. For analytics it maybe interesting to focus only on added and updated? It is rather an easy fix.

[akosyakov]

@loujaybee @andreafalzetti Do you think it would be useful to add an attribute like operation to distinguish what happened, i.e. added, updated, removed? I looked right now at analytics it can be misleading, i.e. you can see 2 modifications to tasks, but final maybe was removing so actually tasks are not used. For analytics it maybe interesting to focus only on added and updated? It is rather an easy fix.

Actually I'm wrong we cannot distinguish removed from updated because our goland types using string types not pointers. It will be quite impactful change, let's leave it for later.

[akosyakov]

/unhold