fatal: [controller0]: FAILED! => { "msg": "Timeout (12s) waiting for privilege escalation prompt: " } about a-universe-from-nothing HOT 9 OPEN

VMs have a larger overhead than containers, so while you might be able to squeeze many tens of them onto a compute node, they will probably not be very efficient. It really depends on what you are using them for.

Typically we use Ceph storage with OpenStack. Up to the Nautilus release we typically use ceph-ansible to deploy it. Then follow this guide to connect it to OpenStack. From the latest Ceph Octopus release you can use the new Ceph Orchestrator to deploy it.

This workshop is really designed to be a tool to learn about using Kayobe. To create a production cluster, you should refer to the kayobe documentation for more information. You can refer back to this set of configuration for a minimal example. What the documentation lacks at the moment however is a complete end-to-end deployment and configuration guide. You can join the #openstack-kolla IRC channel on freenode if you have questions.

from a-universe-from-nothing.

Hi. It looks like you are unable to access the seed machine via SSH, or possibly having problems with sudo. Have you already deployed the seed? Could you try the following:

ssh [email protected] sudo echo hi

Sometimes we see that error on slower hardware. You could try setting export ANSIBLE_TIMEOUT=30.

from a-universe-from-nothing.

from a-universe-from-nothing.

from a-universe-from-nothing.

Which branch of the a-universe-from-nothing repo are you using?

I'm not sure what the NTP problem is. Does the seed have access to the configured NTP servers?

Console logs for the overcloud hosts are in /var/log/tenks. These might help to diagnose any boot failures.

from a-universe-from-nothing.

I'm not sure what the NTP problem is. Does the seed have access to the configured NTP servers?

I've seen the same NTP issue in several other environments. I assume this is caused by firewall restrictions.

from a-universe-from-nothing.

Previously, I was using stable/train branch. Now i am testing stable/ussuri branch. I will update you if the issues resolved. Currently on stable/train the state of each vm is "paused".

Another thing i want to ask as a newbie in openstack (only experienced with VIO), i have enough storage and compute resources available on a vsphere datacenter. I need to ask can i deploy a production level HA openstack hosting maybe 200 to 300 VMs with kayobe? If so, how can i replicate the whole infrastructure i.e 3 controllers, 3 compute and 3 storage nodes with HA supported? Is kayobe supports NSX-T networks? I shall really appreciate your kind response.

from a-universe-from-nothing.

For Ussuri you will need a CentOS 8 host.

I haven't seen the VMs go paused like that.

Kayobe does support HA deployment. You can include multiple hosts in each group in your inventory. NSX-T does claim to integrate with OpenStack (https://docs.vmware.com/en/VMware-Integrated-OpenStack/5.1/com.vmware.openstack.install.doc/GUID-D9BD61E1-E444-441E-90E7-4921A5014CA3.html), although I have not tried it. Typically people use Open vSwitch or Linux bridge as the networking driver.

Are you suggesting to run 200-300 VMs on 3 compute nodes?

from a-universe-from-nothing.

Thanks for your valuable insights
Actually i am not sure, how much workload a compute node can accommodate. Therefore, i am asking you about the best use case scenario in my requirements (200 to 300 VMs). I am well aware of PAAS (Openshift) which can supports 250 pods per compute node, so we can add compute nodes if it is needed. Obviously this type of restriction will also be applied to Openstack. Further i also need to set storage nodes (block, object) how can i do this from your workshop guide? can i use your workshop as a baseline to setup production grade clusters? Thanks once again

from a-universe-from-nothing.