v2.6.1: argo list --chunk-size truncation should apply after other filters

[samath117]

Checklist:

• I've included the version.
• I've included reproduction steps.
• I've included the workflow YAML.
• I've included the logs.

Summary:

argo list takes multiple options which act like filters, like --since [time] or --completed. It also provides an option --chunk-size (default 500) which truncates the list. However, these two operations happen in the wrong order: First the list is truncated (sorted alphabetically), then the list is filtered. Since the purpose of --chunk-size is to limit the size of the output, these two should happen in the opposite order.

In my situation, I discovered this entirely by accident, and it took me a while to figure out what was going wrong. I exceeded 500 workflows in a namespace, and then argo list started silently dropping my workflows that were after the 500th alphabetically. I obviously eventually figured this out, but it was a really bad UX to have argo list just stop working on me.

A simpler alternative might be for --chunk-size to have a default of 0, so users like me don't accidentally stumble into this problem in this way.

I'm also not sure that --chunk-size is working the way it should -- I just get as output the first 500, not all of them in chunks. If I'm supposed to be using it differently, there should be better documentation on that.

---

Message from the maintainers:

If you are impacted by this bug please add a 👍 reaction to this issue! We often sort issues this way to know what to prioritize.

[changhc]

I'd like to take this ticket

[thundergolfer]

I'm also not sure that --chunk-size is working the way it should -- I just get as output the first 500, not all of them in chunks. If I'm supposed to be using it differently, there should be better documentation on that.

I agree with this. Just upgraded to 2.6.1 from 2.4.something and think it's not right that --chunk-size has changed behaviour to truncate the list.

Is this issue changing things such that you have to use --continue to get it to return the full list? I would've thought that --chunk-size should keep it's old behaviour, or has kubectl itself changed the behaviour of its --chunk-size option? #2820