Telemetry does not catch all errors due to abrupt os.Exit calls

[valaparthvi]

/kind bug

What versions of software are you using?

Operating System:
Fedora 34
Output of odo version:
odo v2.2.4 (0c0aaa5)

How did you run odo exactly?

1. odo delete out of a context directory.
2. The above command will fail.

Actual behavior

No telemetry data received on Segment, because at one point during this command execution, LogErrorAndExit is called, which exits before the telemetry data request can be triggered.

I noticed this with odo create or basically any command where you don't have the permission to create comp in a project.

On tracking down, it seems that this is the case with all the command execution where LogErrorAndExit is called at some point, except in runnable.go.
See - https://github.com/openshift/odo/search?q=LogErrorAndExit

Expected behavior

Telemetry data must be sent for these errors.

Any logs, error output, etc?

I can think of 2 solutions, atm:

1. We could either practice a strict behaviour where LogErrorAndExit can only be called by the runnable.go:GenericRun
2. Or, we could put the Telemetry data trigger request inside LogErrorAndExit, which would mean modifying LogErrorAndExit and all of its usages to be able to pass data that is required for triggering the request.

[valaparthvi]

Creating a component in an already existing component directory is another example.

[dharmit]

IMO, this issue is more about getting rid of abrupt os.Exit calls than Telemetry itself.

[anandrkskd]

In function LogErrorAndExit() we are using os.Exit(1) call, And this function is used in multiple places. So do we want to remove os.Exit from this fuction too?

[dharmit]

In function LogErrorAndExit() we are using os.Exit(1) call, And this function is used in multiple places. So do we want to remove os.Exit from this fuction too?

That's actually the right place for os.Exit(1) to exist. If you see, it's getting called in the runnable.go many a times: https://github.com/openshift/odo/blob/e347c7ac0e22ca370ab5df74cd56f4830cb885c9/pkg/odo/genericclioptions/runnable.go#L81-L99

Try to figure out calls to os.Exit in the pkg/odo/cli layer first. And also pkg/ layer but sans pkg/odo/genericclioptions.

Once you have those calls, try to figure out if they should be there or should they be returning errors instead so that we could get data in telemetry. I think most such calls could be replaced by returning an error instead.