Very high quality content. How is this kolla evolving ? Is it still worth to lean ? What direction do you recommend in 2020 ? OSA ? Thanks alot Robert !
Kolla seemed like the way forward when we posted this tutorial and the Kolla Ansible project is likely the way forward if you appreciated the original approach: opendev.org/openstack/kolla-ansible (ps. We've been Ansible fans for a long time, so... biased opinion FWIW. 😉)
Excellent tutorial, thank you! Any plans on updating the tutorial GitHub repo with deployment options beyond Packet, such as VM-based (EC2, GCE, Azure) and other bare-metal-based (MAAS)?
I have some code for a Digital Ocean (also launched via terraform) based system. I felt there was too much "how to launch stuff in AWS" for it to make sense, but I certainly can look at that. In most cases though, the only " tricky" bit of _any_ cloud deployment is getting a useable base OS (the initialize/configure_baseline style ansible helps there), and getting an inventory file that works. With MAAS, the same sort of effort applies.
I appreciate your prompt reply. Should you have a chance to expand the tutorial in that direction, I would be interested in seeing it. I think that a public cloud-based OpenStack might be useful for learning and/or pilot implementations, but most real OpenStack projects IMHO make sense on a private infrastructure, hence my interest in MAAS or similar approaches.
Aleksandr, I've been working with the Digital Rebar Provision (DRP) engine (rebar.digital) for MAAS like functions, as yes, a deployment in Packet, etc. only really makes sense if you want to use the packet infrastructure for VM workloads, and for other public clouds it makes even less sense. Bare Metal private is where OpenStack is intended to be used, or as a basis for a service provider cloud of some nature. I may well build a DRP->OS Cloud on bare metal tutorial in the near future.
Thank you, Robert. I am aware of Digital Rebar, as I try to follow progress in the cloud infra space. I am glad that you agree with me on core ideas. In regard to Digital Rebar, I am not a big fan of their approach of creating a separate OSS project outside of the OpenStack and Kubernetes ecosystems. I think that it would be much better, if they would join and improve relevant existing projects, such as Kolla. Just my 2c.
I think there are different tools for different parts of the space. DR/DRP fills the "how do I go from bare metal to infrastructure via code" part of the equation, much like Ubuntu MaaS, or I guess Foreman{?} from RedHat. Sure you could go Triple-O and use Openstack to manage your bare metal with MaaS, but you then have a system that owns everything, and in some ways is pretty heavy for just the bare metal provisioning piece. Kolla is great for dealing with the "how do I go from code to OpenStack" question, and starts to address some of the day 2 aspects as well (such as n+1 upgrades, and at least some level of log centralization). Kubernetes still needs much of that infrastrucutre (KubeSpray + Prometheus is perhaps the closest/most similar). But there's still a gap in the day 0 story in my view, which is where I tend to find great value in DRP.
i did multinode deployment of openstack using kolla-ansilbe. everything got deployed. everything is fine. but am not able to boot a image in the comptue node. when i check nova-scheduler log it says no valid hosts available. but when i check with nova hypervisor list it shows my compute node. but its able to do boot the image in te controller. the error occurs when i try in the compute node. please help me out in this
Sri, this is usually due to a mismatch in the hypervisor type: kvm vs. qemu. If you are deploying in a virtual environment, often you need to explicitly call out qemu as the hypervisor type in order to allow virtual machines to launch. Have a look at lines 40-48 of this script: github.com/kumulustech/kolla-multinode/blob/master/debian-network.sh This will set the hypervisor type correctly for kvm vs. qemu systems.
Kumulus Tech it taking expressions of interest for a full OpenStack Kolla course - want to learn more and stay in the loop? Check this link: kumul.us/openstack-kolla-course/
Very high quality content. How is this kolla evolving ? Is it still worth to lean ? What direction do you recommend in 2020 ? OSA ? Thanks alot Robert !
Kolla seemed like the way forward when we posted this tutorial and the Kolla Ansible project is likely the way forward if you appreciated the original approach: opendev.org/openstack/kolla-ansible (ps. We've been Ansible fans for a long time, so... biased opinion FWIW. 😉)
Excellent tutorial, thank you! Any plans on updating the tutorial GitHub repo with deployment options beyond Packet, such as VM-based (EC2, GCE, Azure) and other bare-metal-based (MAAS)?
I have some code for a Digital Ocean (also launched via terraform) based system. I felt there was too much "how to launch stuff in AWS" for it to make sense, but I certainly can look at that. In most cases though, the only " tricky" bit of _any_ cloud deployment is getting a useable base OS (the initialize/configure_baseline style ansible helps there), and getting an inventory file that works. With MAAS, the same sort of effort applies.
I appreciate your prompt reply. Should you have a chance to expand the tutorial in that direction, I would be interested in seeing it. I think that a public cloud-based OpenStack might be useful for learning and/or pilot implementations, but most real OpenStack projects IMHO make sense on a private infrastructure, hence my interest in MAAS or similar approaches.
Aleksandr, I've been working with the Digital Rebar Provision (DRP) engine (rebar.digital) for MAAS like functions, as yes, a deployment in Packet, etc. only really makes sense if you want to use the packet infrastructure for VM workloads, and for other public clouds it makes even less sense. Bare Metal private is where OpenStack is intended to be used, or as a basis for a service provider cloud of some nature. I may well build a DRP->OS Cloud on bare metal tutorial in the near future.
Thank you, Robert. I am aware of Digital Rebar, as I try to follow progress in the cloud infra space. I am glad that you agree with me on core ideas. In regard to Digital Rebar, I am not a big fan of their approach of creating a separate OSS project outside of the OpenStack and Kubernetes ecosystems. I think that it would be much better, if they would join and improve relevant existing projects, such as Kolla. Just my 2c.
I think there are different tools for different parts of the space. DR/DRP fills the "how do I go from bare metal to infrastructure via code" part of the equation, much like Ubuntu MaaS, or I guess Foreman{?} from RedHat. Sure you could go Triple-O and use Openstack to manage your bare metal with MaaS, but you then have a system that owns everything, and in some ways is pretty heavy for just the bare metal provisioning piece. Kolla is great for dealing with the "how do I go from code to OpenStack" question, and starts to address some of the day 2 aspects as well (such as n+1 upgrades, and at least some level of log centralization). Kubernetes still needs much of that infrastrucutre (KubeSpray + Prometheus is perhaps the closest/most similar). But there's still a gap in the day 0 story in my view, which is where I tend to find great value in DRP.
i did multinode deployment of openstack using kolla-ansilbe. everything got deployed. everything is fine. but am not able to boot a image in the comptue node. when i check nova-scheduler log it says no valid hosts available. but when i check with nova hypervisor list it shows my compute node. but its able to do boot the image in te controller. the error occurs when i try in the compute node. please help me out in this
Sri, this is usually due to a mismatch in the hypervisor type: kvm vs. qemu. If you are deploying in a virtual environment, often you need to explicitly call out qemu as the hypervisor type in order to allow virtual machines to launch. Have a look at lines 40-48 of this script: github.com/kumulustech/kolla-multinode/blob/master/debian-network.sh This will set the hypervisor type correctly for kvm vs. qemu systems.
Kumulus Tech it taking expressions of interest for a full OpenStack Kolla course - want to learn more and stay in the loop? Check this link: kumul.us/openstack-kolla-course/
Excellent. Thank you very much for uploading. I was able to install and configure OpenStack using kolla-ansible but when I create an instance I can't seems to login or ping those instances. Here is my global.yml file:
kolla_enable_sanity_checks: "yes"
kolla_base_distro: "centos"
kolla_install_type: "binary"
openstack_release: "pike"
node_custom_config: "/etc/kolla/config"
kolla_internal_vip_address: "10.50.50.17"
kolla_internal_fqdn: "{{ kolla_internal_vip_address }}"
kolla_external_vip_address: "{{ kolla_internal_vip_address }}"
kolla_external_fqdn: "{{ kolla_external_vip_address }}"
network_interface: "br0"
kolla_external_vip_interface: "{{ network_interface }}"
api_interface: "{{ network_interface }}"
storage_interface: "{{ network_interface }}"
cluster_interface: "{{ network_interface }}"
tunnel_interface: "{{ network_interface }}"
dns_interface: "{{ network_interface }}"
neutron_external_interface: "br1"
neutron_plugin_agent: "openvswitch"
openstack_logging_debug: "False"
nova_console: "novnc"
enable_horizon: "yes"
enable_neutron_provider_networks: "yes"
enable_magnum: "yes"
enable_mistral: "yes"
enable_tempest: "yes"
nova_compute_virt_type: "kvm"
tempest_image_id:
tempest_flavor_ref_id:
tempest_public_network_id:
tempest_floating_network_name: