-
Notifications
You must be signed in to change notification settings - Fork 95
Open
Description
Issue
Creating a cluster with civo kubernetes create
creates a cluster with defaults. This is phenomenal use of convention over configuration so the user doesn't need to input too many flags or commands to create a cluster.
However we should perhaps reconsider the defaults used here, right now this is what it does:
- Create a cluster with 3
g4s.kube.medium
nodes --> We should probably make theg4s.kube.xsmall
the default. 3 nodes may be okay, although perhaps 2 is more conservative. - Randomly pick a name for the cluster --> This is phenomenal, no need to change
- Randomly pick a Linux distro for cluster --> I don't think this is a good thing as it changes the nature of the cluster too much.
- Create a firewall/s for the cluster --> This is also okay, but cluster deletion should clean these, or at least provide an option to do so. Will create another ticket for this.
(Not sure if I am missing other defaults)
In addition, the UI should be aligned with the defaults the CLI provides.
Acceptance Criteria
- Reconsider changing the defaults to create a cluster that's as cheap as possible for generalised workloads.
- Make the OS deterministic, don't randomize it.
Cluster defaults should be
- As secure as possible.
- As cheap as possible.
- Catering to general needs of the public wishing to test a cluster or use for quick, simple workloads. We perhaps shouldn't assume that the defaults are for heavy workloads, or even for production workloads as in those scenarios a user may want more fine grained settings
- The nature of the cluster should be predictable, randomizing the OS prevents that.
Another option could be to create templates for different types of configs, for example:
- ML
- Production
- Test
- Budget
- Etc
But this is out of the scope of this ticket.
Metadata
Metadata
Assignees
Labels
No labels