gecos-team/gecosws-config-assistant

Workstation recipes are duplicated in GCA

Closed this issue · 8 comments

alfem commented

GCA package includes a copy of workstation recipes, and sometimes these are out of sync.

If we can demonstrate this copy is not necessary, it must be deleted.

I propose to split gecos-workstation-management-cookbook in two: one for GECOS Control Center policies, and one for GECOS Config Assistant recipes. This one will be the only one shipped in GCA package.

It seems that GCA uses the following recipes:

  • local (includes sssd_mgmt)
  • misc_mgmt
  • network_mgmt
  • unlink_from_sssd
  • unlink_from_gcc
  • unlink_from_chef

So the following recipes aren't used:

  • broadband_mgmt
    • default (includes "software_mgmt", "misc_mgmt", "network_mgmt", "users_mgmt" and "printers_mgmt")

But, since "default" recipe includes "misc_mgmt" and "network_mgmt", if you want to split the cookbook in two GECOS will have two copies of both recipes in both cookbooks.
Are you OK with this?

Keep in mind that GCA can be run at any time, not always in an uninitialized node (you can run GCA to unlink from GCC or to change authentication); I mean we can not rely on recipes not to be run due to lack of corresponding data in node json.

alfem commented

Could you point at the GCA code whre misc_mgmt and network_mgmt are used?

Perhaps it is unnecesary.

"misc_mgmt" is used in gecosws-config-assistant/firstboot/serverconf/init.py to create a chef-solo script to link to the GECOS Control Center. And "network_mgmt" is used in the same file to create a chef-solo script to configure the client authentication method.

Both scripts seems to be the core of GCA.

Hi all!

I've sent a pull request (#11) that fixes the problematic by building the gca package with the latest upstream cookbooks (gecoscc-chef-repo master branch); matching this way, the cookbooks used by the gecoscc.

alfem commented

Could you make the PR against development?

Trusty is the 'master' in these distro packages.

Done (#12)!

I just realized that there was an error initializing git sub-subdmoules, so I sent a new PR (#13).

Bests!