search

ketzacoatl / compute-demo

HA compute / data processing cluster on AWS, built with Terraform, Packer, Nomad, Consul, and SaltStack.
3 stars 0 forks source link

rename "manage" node class to something better #15

Closed ketzacoatl closed 6 years ago

ketzacoatl commented 6 years ago

master is out, so is manage. ops? support? admin? ops-workers? non-compute?

kerscher commented 6 years ago

If there is only one cluster, why not compute-demo?

ketzacoatl commented 6 years ago

sorry, within the demo as a whole, there are 4 different roles. The roles are bastion, management, compute, and load-balancer. In nomad we map the role to "node classes", though nomad does not run on the bastion, so there are only 3 node classes. The "Management Tier" has the name "manage" in the asg, I am looking for a better word/phrase there. The purpose of the "Management Tier" is to run the ops tooling that is used/needed to run/manage/operate the deployment/env/cluster as a whole. So Prometheus, grafana, autoscaler, dns, logging tools, security tools, other stuff that is ops/admin in nature.

kerscher commented 6 years ago

manage does not strike me as bad at all. You could in theory use something like system, or admin I guess, but I don’t see a problem with the current one.

ketzacoatl commented 6 years ago

I think I would vote as compute and ops as a first choice, and maybe compute and sys as a second choice.

kerscher commented 6 years ago

After more thought and reading, it seems system or sys would be confusing, since system scheduled jobs run on all nodes. Between ops and manage, I actually prefer manage.

ketzacoatl commented 6 years ago

OK, I'll agree to leave it as-is for now.