jay7x / puppet-lima Goto Github PK

It'd be nice to have a VM startup/shutdown order in the cluster. That allows to do staged cluster startup/shutdown.

Imagine we have 3 VM with node1 having order of 10 and other 2 VMs having order of 100. Then node1 should be started first and the other nodes should be started after the 1st one in parallel. Shutdown sequence should happen in a reversed order.

lima::resolve_reference task uses limactl show-ssh to get a VM ssh config. The command output has multiple IdentityFile lines which makes it impossible to use as-is in the Bolt inventory. Currently the value is hardcoded as "#{lima_home}/_config/user".

Starting from Lima 0.15.0 (I don't remember it in earlier versions at least) there is IdentityFile field in the limactl list --json output for every VM. So I guess it's better to use this value instead of hardcoded one as we have full JSON object from the limactl list output anyway.

Another frequent nice-to-have feature is cluster restart. Options:

1. Specific plan lima::cluster::restart which calls stop and start inside
2. restart=true option in the lima::cluster::stop plan

While starting multiple VMs in parallel is unsafe in general (see lima-vm/lima#1354) it'd be really nice to have a way to 🤞🏻, 🙏🏻 and do it anyway. As long as all VMs are created already it should be safe to start them in parallel. So my idea is to add a parallel=true kind of parameter to the lima::cluster::start plan.

Starting from some lima version (0.17.0) limactl create command was implemented. It'd be nice to use it instead of guessing what is expected to do in a start call.

It'd be nice to validate the config hash passed to the lima::start task if defined.

N.B.: There is a way to pass config hash to the limactl start via stdin.. this way we don't need to create a temporary file => we cannot check it with limactl validate..

It'd be nice to have a way to configure that a VM is plain. I believe it should be possible to define this in a YAML config. Otherwise another parameter is required for the lima::create task (and for the lima::cluster::create somehow).

As limactl delete allows to specify multiple VMs to delete it'd be nice to allow this in the lima::delete task.

It'd be nice to use Lima cli helper library instead of calling it explicitly from every task. I implemented such a helper in the beaker-lima project. It might require some tuning though.

With the separate lima::cluster::create plan introduced recently it'd be nice to have an ability to create and run a cluster in a single plan call:

1. lima::cluster::create start=true 👍🏻
2. lima::cluster::start create=true 👎🏻

As lima has no way to define a cluster it'd be nice to have few plans to define/create/delete multiple VMs as a "cluster" entity.

E.g. one may want to run local k8s cluster in lima (N master nodes, M worker nodes e.g.). Another example is local Puppet cluster (1 Puppetserver with CA, 1 PuppetDB, X agent nodes).

The info task should take no args and return the limactl info result

Quite frequently I'm stopping a cluster to delete it after. So it'd be nice to have ability to stop a cluster and delete it in the same plan call. There are options:

1. Add delete=true parameter to the lima::cluster::stop plan
2. Add stop=true parameter to the lima::cluster::delete plan
3. Implement both options above