zikalino / openshift-work Goto Github PK

Here's an example of creating VMSS / scaling up / deploying application.

https://github.com/Azure-Samples/ansible-playbooks/tree/master/vmss

I think this should be done once we are done with initial implementation
However we should find another way of deploying scripts as VM Extension operates on VMs directly.
Perhaps it would be possible to do it on VMs created by VMSS, but it doesn't sound like an ideal solution.

Actually, i have another idea how this could be done:

• we could have initial playbooks to just create and store images
• main script would just create 3 scalesets
• scaling up wouldn't require any additional steps on new vms

While we develop the template, we can skip the VM extension provisioning to save time during CI/CD runs.

Make sure to have a proper structure that will allow us to keep track of all the failed activities while minimizing the provisioning time

currently username and password is used

Let's try to do it all in Ansible Playbooks

Create a bastion VM ready to be used

We need some other way to pass the variables from the first play to the second host.

Currently it fails here:

 # $4 MASTERPUBLICIPHOSTNAME=$4 - parameters('openshiftMasterPublicIpFqdn')
"{{master_publicip_fqdn }}" \

Because it doesn't have master_publicip_fqdn

Talking with @haroldwongms, we saw that the deployment time of openshift playbook on a fully provision infrastructure is of 1hr 07 minutes, even after running the prep scripts.

This issue is a call for heros that want to revise that legendary playbook:
https://github.com/openshift/openshift-ansible/blob/master/playbooks/deploy_cluster.yml

And the script @haroldwongms is using to deploy it:
https://github.com/Microsoft/openshift-origin/blob/master/scripts/deployOpenShift.sh

And see if we could customize or pass some flags to reduce the deployment time.

Please check task: "Run deployOpenShift.sh"
And make sure to add all the correct parameters to run it.

Stop using username/password

I have noticed that if something goes wrong while running tasks asynchronously, there will be no information displayed at all, so user may be left with an impression that everything went fine, but some resources are not really created.
This happened to public ip addresses, and virtual machines while testing.

Make sure that there are no hardcoded elements in the playbook and provide an easy way of creating multiple clusters without collisions.

I have managed to create and dump infrastructure using ARM template.
It also creates CNS nodes, every CNS node has at least 4 disks attached:

• docker_pool
• gluster_disk 0 … 2
• os_disk
  I don't understand the meaning of these nodes, they are not mentioned in the documentation in ARM repo.

Dynamically we will have to:

• Provision a bastion VM - #13 by @asinn826

• Create a new pair of SSH keys

• Propagate the SSH keys to the other VMs with authorized_keys module

• Add it to the ansible host list so it can perform actions like adding running the Openshift provisioning playbook and running verifications on the other VMs

   - name: Add new instance to host group
        add_host:
          hostname: "{{ item['ips'][0].public_ip }}"
          groupname: azure_vms
        with_items: "{{ azure.deployment.instances }}"

Right now, everything is provisioned correctly. But we still need to make sure that the VMs are loaded and ready to be used.

Need to make sure that all the VMs are up and running and that SSH is ready to go.
Specifically, we need to wait for the bastion to be ready and run the deployment from there

I have just been testing whether NAT is configured propely, and seems like it's not.

I have tried to add inbound_nat_pools but it didn't have any effect.

I also tried to create NAT rules using natpool_ but it also had no effect.

I think existing azure_rm_loadbalancer module may be not working correctly

Ansible should create resources with the maximum parallelism possible

Break the playbook into pieces and make sure that they have no dependencies to minimize deploying time.

Add to the create yaml file the code for nic, vms, AS