search

open-infrastructure-labs / ops-issues

0 stars 0 forks source link

Figure out shortest path between new cluster install and using ArgoCD for cluster management #4

Open larsks opened 3 years ago

larsks commented 3 years ago

Current cluster configuration has grown organically. How do we minimize the amount of configuration necessary to get from an initial install to having argocd running?

HumairAK commented 3 years ago

After the re-installation of the zero cluster we got a better idea of the path from no cluster, to a cluster provisioned via ACM and installing our argocd tooling on top of that afterwards.

After ACM installation, we simply needed to manually run kustomize build upon our argocd manifests repo found here. Not having the post-install operators install did not seem to pose as an issue.

ooichman commented 3 years ago

@HumairAK , do you want to build an Anisble Operator as a wrapper for kustomize so when you create a new CR it will run ?

tumido commented 3 years ago

@ooichman that just shift the workload - instead of running kustomize build | oc apply you would do ansible-playbook I don't see a difference.

tumido commented 3 years ago

What is the exact action item to make this complete? Thoughts?

HumairAK commented 3 years ago

I think we should re-review the steps taken during the cluster migration, and what was required to go from acm provisioned zero cluster to a state where argocd is deployed and store this in a doc. I'm thinking a runbook. Doesn't have to be super detailed, WDYT?

larsks commented 3 years ago

Now that we're running ACM and ArgoCD on the moc-infra, cluster, what are the steps beyond "install the target cluster using ACM"? I assume it's mostly argocd configuration.

tumido commented 3 years ago

The steps to set up ArgoCD to manage clusters installed via ACM are exactly this: https://access.redhat.com/documentation/en-us/red_hat_advanced_cluster_management_for_kubernetes/2.2/html-single/manage_applications/index#configuring-argo

Digest is: When provisioning use:

apiVersion: agent.open-cluster-management.io/v1
kind: KlusterletAddonConfig
...
spec:
  applicationManager:
    argocdCluster: true

That's it. The cluster is then directly available to be deployed to in ArgoCD. No additional steps are needed.

larsks commented 3 years ago

That's kind of what I was getting at. Is this ticket resolved by the fact that we're using externally hosted ACM and ArgoCD?

tumido commented 3 years ago

Not sure... I think we don't have all the same understanding about the ticket.

If it is "what should we do once we provision a cluster to get it into ArgoCD" then yes, the fact that we use ACM basically resolves the ticket.

If it is "how to get ArgoCD running on the first cluster in an environment (like infra in our case)", then we need the runbook @HumairAK mentioned that would say do oc apply with this resources.

HumairAK commented 3 years ago

I think I may have muddled the water by mentioning the migration since that was only pertaining to the zero cluster and not management. I think documenting both is important. What do we think about the action items for this issue being:

?