TODO items for ansible commander

4/2 NOTES

supervisord to start celery, modify ansible playbook to set up supervisord <- ChrisC
make job rest triggerable & job status readable. now happens just by creating a new job <-- MPD
do we need something other than default playbook (ProjectOptions) <-- me, later
way to send cntrl-c to kill job (method on job?) <-- ChrisC, low priority
documentation on how to run with callbacks from NOT a launchjob <-- ChrisC
interactive SSH agent support for launch jobs/creds <-- ChrisC
michael to modify ansible to accept ssh password and sudo password from env vars

CH notes

need to add api/v1/group/N/all_hosts need to add api/v1/host/N/all_groups or make sure it's in the resource figure out how to mark each object generically w/ whether it can be edited/etc enable generic filtering sorting? if you delete a subgroup, hosts go up to parent verify unique_together on inventory and host name

make a way to start a job via the API (create a new job)

future feature of being able to create job templates -- supported in db by JobTemplate

REST TODO

credentials objects & permissions
tags (later)
audit trails (later)
jobs triggering <-- important
related resources on everything that makes sense <--- important
expose log data from callback (decide on structure)

api/v1/ <-- discoverable api/v1/me

{ id: 2 foo: 3 related: [ 'bar' : 'http://api/v1/users/5/bar', 'baz' : 'http://api/v1/users/10/baz', } }

LATER

UI layer, CLI client (and libs)
clean up initial migrations
init scripts (supervisord), Apache proxying, etc
does inventory script need any caching (??)
support multiple project launch options (different flags, etc)

TWEAKS/ASSORTED

security check - playbook must be fully pathed when executed by commander and relative to project directory. Project directories must be unique and be (also no ../, etc). default_playbook MUST be a relative path (no starting with . or / or containing ..).
when creating a project, write a '.acom' file, this file contains a JSON list of valid project GUIDs that are allowed to use it. This can be created automatically by the system if it does not already exist, but not overwritten by the system.
document multi-org content limitations with regard to access security and sensitive vars
when app starts it creates a default organization and all superusers are automatically added to it
API discovery resource at /api and /api/v1
audit/test read only fields like creation_date
ability to import inventory from script (adds variables and merges, does not override, does not remove anything)
ability to import inventory from inventory files (upgrade case)

QUESTIONS

if creating a project, do we want to have an appliance style path for them, like /storage/projects/GUID ??? may want to keep somewhere else ?

MISC

when associating inventory objects, objects should share common inventory records (user_can_attach method should check)

PAGINATION, FILTERING, SORTING -> CH

3.1 KiB Raw Blame History