inconsistent/unusable exit status in argo CLI wait

[Satook]

Hi there,
I'd like to submit a PR for a --result-in-exit-status feature for the CLI argo wait command.

After reading the code at https://github.com/argoproj/argo-workflows/blob/main/cmd/argo/commands/common/wait.go
argo wait will exit 1 if:

1. the workflow cannot be found
2. there's an error establishing the wait
3. the workflow finished in the Failed or Error phase

I'm using argo wait as part of a multi-cluster workflow.
This makes the exit code essentially useless for determining if I need to retry the wait step or not.

Instead I'm capturing stdout and doing a container exit of -1, 0, 1 so the retryStrategy.expression can retry on '-1'
With the following shell logic

if [ -z "${result}" ]; then
    # no output -> transient failure
    exit -1
fi
if [[ $(echo "${result}" | cut -d ' ' --fields 2) == "Succeeded" ]]; then
    # confirmed workflow success
    exit 0
else
    # confirmed workflow failure/error
    exit 1
fi

Would anyone object to me making the exit status of argo wait more explicit behind a (default off) flag?

Context

For context I have an overall CI/CD workflow that has to run sub-workflows on other clusters as part of its validation and pre-deployment checks and preparation (think smoke tests, projection rebuilds, etc).

I can't just use a workflow on 1 cluster within a single namespace.

I run the "other cluster" workflow in 2 steps

1. Step 1: kick off using argo submit to capture the workflow name
2. Step 2: use argo wait with output "parsing" so my "parent" workflow can capture the overall success/failure/error status of that workflow
   • with retries to work around transient issues, e.g. connection loss, firewall hiccups, controller restarts, etc.

Cheers,
Andrew