Re: Resources offer: Catalyst Cloud
On 12/22/18 11:57 AM, kuLa wrote:
> On 2018-11-13 13:57:18, Andrew Ruthven wrote:
>> Hey,
>
> Hi Andrew,
>
>> The Catalyst Cloud would like to offer the Debian Cloud Team resources
>> on our OpenStack based public cloud. Please let me know if this is of
>> interest and we can work out what it'd look like.
>>
>> https://catalystcloud.nz
>
> Sorry for being late to the subject but I think that this hasn't been addressed
> yet so here are my 2p.
>
> Thanks for the offer.
>
> I know that we (as Debian) have access to AWS and GCP. We also have a kind of
> access to Azure tho this arrangement as far as I know this is not directly
> arranged between Debian and MS.
>
> I don't know about any Openstack installation available to Debian as a project
> for testing our images or just to have a play with it. So IMO we should
> consider this offer and maybe use it as a part of our CI pipeline.
Well, more than a year ago, in Seattle, I installed a self-hosted
OpenStack in an all-in-one VM, so we could use a script to validate the
OpenStack image. It was running the tempest script. This is normally
enough to run all the tests we need on the OpenStack image.
Unfortunately, even if the script was super simple to use, it never was
used. My understanding since the last meeting, is that there's a Unix
right problem with the VM running under the wrong Unix group. It's
amazing it took a full year so that I got the info, but hey, let's fix
that...
Now, an external resource which we don't control isn't going to change
anything to the problem I just mentioned. Even worse: I don't see a
public cloud provider making tempest available to us, as it would make
their cloud dirty (ie: with maybe some remaining aftefacts after the
script runs, some nasty tempest configuration that needs to be
installed, etc.).
Now that I'm kind of forcibly loosing the sole control of the image
(which could have been a good thing), anyone can break things in the
OpenStack image without any CI/CD result to warn us about it. It's
really not going in a direction where I want to put proudly say I was
involved or something. This needs to change, and the change must happen
rapidly, as Buster is getting frozen soon.
So, this question goes to both the DSA and to Steve: what do I need to
do so that the tests I proposed in 2017 can effectively be run on each
build of the OpenStack FAI image? Can we solve this problem soon?
Cheers,
Thomas Goirand (zigo)
Reply to: