The playbook can now optimize itself based on the enabled components in
for all hosts in the inventory (`just optimize`) or for a specific host
(`just optimize-for-host HOSTNAME`).
The optimized playbook will have:
- fewer requirements (fewer roles need to be installed by `just roles`)
- a shorter and quicker to evaluate `group_vars/mash_servers` file
- a `setup.yml` file which includes less roles
Running the playbook optimized is still work in progress.
There still probably exist various role dependencies in the group-vars file, etc.
The `optimize-reset` command aims to restore your playbook to a
non-optimized state, which should work as before (and not experience bugs).
The playbook takes care to notice of changes to the various files in
`templates/` (`setup.yml`, `requirements.yml`, `group_vars_mash_servers`)
and update your optimized or non-optimized copies that are derived from
these templates. To do this, it keeps `.srchash` files in the `run/` directory.
When it notices a change in the source file's hash (by comparing to the `.srchash` file),
it will update you to the new template.
Optimization state is stored in a file in `run/` as well (`optimization-vars-files.state`).
Should the playbook notice changes in the source `template/` files, it
should update you and re-optimize using the same parameters as before (read from the state file).
In the future, we'll also have optional optimization steps, which could
trim down `setup.yml` based on the components being used.
Related to e2132a3c51 which did the same
for the `requirements.yml` file.
This way, something like `just install-service postgres` will
definitely do a `restart` and not a `stop` + `start` (which may leave
dependant services stopped).
This has potentially been addressed by
7acb5f4b85
as well.
