CentOS 8 Anaconda requires a repo called “AppStream”, which is what is failing for you right now. I created the following repos using hammer scripting (I’ve removed the CentOS 7 repos for clarity). So, AppStream shows up twice, just so I can remove the workaround later. Note, you have to sync the kickstart and os repos separately.
Create a content view presenting all of those to the client systems, and you should be good to go. It’s best if you have activation keys created as well, with all the necessary repositories and Content view associated.
If you plan on using Internet repos, you will need to create a custom Installation Media source with the URL of http://mirror.centos.org/centos/$major/BaseOS/$arch/kickstart
I would also suggest perusing the documentation sites for additional information, as they are getting more detailed in current releases.
We are still advising people to follow the CentOS 8 instructions that @Mkbrown noted in the Provisioning Guide. The specific instructions are due to a bug in Anaconda for CentOS 8 that looks like is still around: https://bugs.centos.org/view.php?id=16470
Otherwise it should be working, please let us know if not.
I am afraid that the comment “it’s probably Foreman issue” indicates this might be a long one. Is there a workaround we could implement in Katello or somewhere else for the time being?
I believe the cleanest way to fix this is to implement a custom code in Pulp for CentOS8 - it needs to detect CentOS8 and set up repositories in a way that it is compatible with Anaconda.
Is the issue that this workaround with the custom kickstart template is too hacky? I haven’t heard yet that these instructions are out of date, unless there was a regression that we’ve missed.
I do agree that this should work out of the box though.
Hmmm I haven’t realized we have that. While this is not ideal, that would work I guess. The OP @hiero-nymus is having issues with Satellite 6 tho, it probably missed 6.6. So can I assume that 6.7 will be fixed?
The AppStream code is older than Katello 3.12 so it might be in there. @hiero-nymus can you try specifying a custom kickstart provisioning template like the instructions say here? Foreman :: Plugin Manuals
As an FYI, the Foreman 1.24.2/Katello 3.14.1 releases do not include the necessary workaround in the ‘Kickstart default’ template, nor in the Installation Media configurations. Those might need to be added to the Community templates, or some special notes created around CentOS 8 provisioning, including all the necessary changes (OS configuration template assignments, Installation Media URL, Repos and Content Views).
I’ll take a look at where we can add more support for this workaround. It would be nice if this can get fixed in Anaconda or Pulp so extra steps aren’t required by Katello users.
@hiero-nymus a couple more questions that might help:
Are your AppStream and BaseOS repositories published in a content view that you’ve set up? If so, they should be same environment and content view. If you’ve only synced them but haven’t published them in a content view you should be all set there.
Is your AppStream repo literally named “AppStream”?
Hi @iballou and sorry for the delay.
Yes AppStream and BaseOS repositories are on the same content view and were published too.
Thanks @Mkbrown, I’ve created a new AppStream repository as you described and follow the workaround’s installation from your link. Actually I’ve started the installation and until now, all is fine.
I just have a comment about the filesystems type. The installation seems to work only with ext4 because I’ve received this error with xfs : something like raise FSError : failed to mount.… As someone an Idea about that?
Not sure if this will help anyone else but I am too impatient to wait for a real fix so here’s what I did:
I was getting the “…/…/…/AppStream/” error just like everyone else.
I read the googles.
Came up with bunk.
Impatience commenced.
Created /etc/httpd/conf.d/pulp_aliases.conf on my foreman server with the following contents:
Alias /pulp/repos/home/Library/BaseOS/x86_64/os /var/www/pub/yum/http/repos/home/Library/CentOS_8/custom/CentOS_8/CentOS_8_-_Base-x86_64
Alias /pulp/repos/home/Library/AppStream/x86_64/os /var/www/pub/yum/http/repos/home/Library/CentOS_8/custom/CentOS_8/CentOS_8-AppStream-x86_64
Alias /pulp/repos/home/Library/centosplus/x86_64/os /var/www/pub/yum/http/repos/home/Library/CentOS_8/custom/CentOS_8/CentOS_8-_centosplus-x86_64
Alias /pulp/repos/home/Library/Extras/x86_64/os /var/www/pub/yum/http/repos/home/Library/CentOS_8/custom/CentOS_8/CentOS_8-_Extras-x86_64
Alias /pulp/repos/home/Library/PowerTools/x86_64/os /var/www/pub/yum/http/repos/home/Library/CentOS_8/custom/CentOS_8/CentOS_8-_PowerTools-_x86_64
Note: you may have to tweak the aliases to match the names of your repositories. Once this is all fixed by whomever, you can delete the pulp_aliases.conf file but it shouldn’t hurt anything even after it is fixed.
We’ve had several reports of difficulties provisioning with CentOS 8 in the last number of weeks. @lzap would it be an idea to look for a ‘volunteer’ to do a walk through of troubleshooting this?