Ask Your Question

bjoernh's profile - activity

2019-10-21 16:22:14 -0600 received badge  Famous Question (source)
2019-09-23 07:10:23 -0600 received badge  Famous Question (source)
2019-09-12 07:09:52 -0600 asked a question Is it possible to skip releases when upgrading with Kolla?

Hi all,

we have a production environment deployed with Kolla-Ansible using the Queens release. We would like to upgrade to Stein, as Queens is running out of maintenance at the end of October.

I have created a testing environment in the cloud resembling the essential parts and configuration of our production environment. I can easily deploy Queens, Rocky, or Stein using Kolla-Ansible. However, when trying to upgrade from Queens to Rocky, results are at least unstable.

The upgrade from Rocky to Stein worked quite well, and I was able to somehow do a 3/4 + 1 step upgrade from Queens to (Rocky) to Stein, where the upgrade to Rocky was about 3/4 successful. However, I am uncertain as to the completeness and quality of this upgrade. The Glance registry was removed, that's a no-brainer. Other than that, everything seems to work at first glance. Heat caused some trouble, which is why I removed it from the upgrade and then enabled and deployed it again afterwards. This would be a viable solution for the production environment, as Heat is currently not used.

Does anyone have any experience with such upgrades? Has anyone tried to skip a release and do a direct upgrade?

Thanks and best regards, Björn

2019-09-12 06:59:28 -0600 received badge  Notable Question (source)
2019-09-12 06:59:28 -0600 received badge  Popular Question (source)
2019-04-02 07:48:17 -0600 received badge  Notable Question (source)
2018-11-13 06:46:08 -0600 asked a question How to enable OIDC behind HAProxy in Kolla deployment

Dear all,

we are using kolla-ansible to deploy OpenStack. At the same time, we have the requirement to integrate with a federation using OIDC. We followed the official OpenStack documentation, but are now stuck at the point where Horizon tries to redirect the client browser to the internal HAProxy endpoint (192.168.x.y), which is not available outside of our cluster. How can we influence the host part of the URL to contain the public facing endpoint?

Best regards, Björn

2018-10-25 07:58:55 -0600 received badge  Popular Question (source)
2018-10-25 07:39:33 -0600 received badge  Student (source)
2018-10-25 04:24:18 -0600 commented answer How do I support additional backend drivers for Cinder in kolla-ansible?

This is exactly what I did. I built my own images using kolla and added the required bits and pieces to kolla-ansible. The images were fairly easy, but rely on a local package repository. The adaptations in kolla-ansible were more involved and deserve a review. Thank you for now.

2018-10-17 01:14:00 -0600 received badge  Enthusiast
2018-09-26 16:28:22 -0600 asked a question How do I support additional backend drivers for Cinder in kolla-ansible?

Hi all,

I intend to deploy OpenStack through kolla-ansible. However, there are a few side conditions that I would like to see fulfilled. We have a GPFS shared filesystem available and would like to exploit it for Cinder, Glance, and Nova. Right now, I am trying to get my head around configuring it through Kolla. However, the deployment guide does not explicitly mention support for GPFS.

Is it even possible now or at least conceivable to support drivers that are not directly supported by Kolla, but are at least generally supported by OpenStack? How would that be done?

I do see a few options here:

  1. direct support through Kolla deployment (my favorite)
  2. Kolla deployment with an external Cinder supporting the GPFS backend
  3. different deployment method (openstack-ansible?)

Regarding option 2, I am uncertain about the implications for the other services, i.e. Glance and Nova.

I am looking forward to any responses or comments. Also open for discussion.

Best regards, Björn

2018-09-26 16:28:20 -0600 answered a question Networking Issues With Kolla-anisble deployment

Hi,

the kolla_external_vip_address is a so called floating IP that changes hosts should the "active" one go down. Such IPs are usually assigned with a netmask of /32. There may actually be another reason why you're unable to access your dashboard. Did you assign the IP to the correct interface, such that it is reachable? Do you have a separate network_interface? Are you doing an all-in-one or multihost deployment.

Best regards, Björn

2018-09-26 16:28:20 -0600 asked a question How do I support additional backend drivers for Cinder in kolla-ansible?

Hi all,

I intend to deploy OpenStack through kolla-ansible. However, there are a few side conditions that I would like to see fulfilled. We have a GPFS shared filesystem available and would like to exploit it for Cinder, Glance, and Nova. Right now, I am trying to get my head around configuring it through Kolla. However, the deployment guide does not explicitly mention support for GPFS.

Is it even possible now or at least conceivable to support drivers that are not directly supported by Kolla, but are at least generally supported by OpenStack? How would that be done?

I do see a few options here:

  1. direct support through Kolla deployment (my favorite)
  2. Kolla deployment with an external Cinder supporting the GPFS backend
  3. different deployment method (openstack-ansible?)

Regarding option 2, I am uncertain about the implications for the other services, i.e. Glance and Nova.

I am looking forward to any responses or comments. Also open for discussion.

Best regards,

Björn