[Elastic Agent] Inconsistency for enroll and install command

[ruflin]

The usage of our enrollment command looks as following:

elastic-agent enroll <kibana_url> <enrollment_token> [flags]

For install it is:

elastic-agent install [flags]

To pass enrollment_token and kibana_url the flags --kibana-url and --enrollment-token are needed but these two flags are missing for enroll. We should be consistent here.

I stumbled over this when I copy pasted the install command but only wanted to enroll so I change install to enroll and it failed as enroll did not have the flags. I like the flags in general as they don't enforce a certain oder so I think we should make the flags also available for enroll to be consistent.

In addition, it should be checked if there are further incosistencies.

[elasticmachine]

Pinging @elastic/ingest-management (Team:Ingest Management)

[blakerouse]

Well I think the inconsistency has validity. That is because for install kibana-url and enrollment-token are optional where as for the enroll command they are required.

[ruflin]

Would it make sense to at least support in enroll to also pass it as flag in addition? Not even sure if this is possible to have both option 🤔

[blakerouse]

It is possible to make the enroll command use those parameters and those parameters be required. Removing the requirement of positional arguments on the enroll command all together.

[ph]

My question would, be does it worth the effort and what we gain? I believe we want to encourage the install workflow? IMHO enroll only makes sense now for testing and even with that, not everything can be tested. Which open actually opens a few more questions.

• Should we keep enroll?
• Should we keep enroll only for testing and hide from --help or special build.

[blakerouse]

You need to keep enroll in the case of DEB/RPM installation, where the install command cannot be used.

[ph]

🤦 oh deb/rpm :(

[ruflin]

If I understand the comment from @blakerouse correctly, we can only move to the params for enroll if we remove the positional arguments. As this is a breaking change, I would delay this.