Red Hat Bugzilla – Bug 978824
Clarify the need for a ftp or http server when importing distros
Last modified: 2016-05-26 09:21:44 EDT
Description of problem:
Currently it is not crystal clear that importing from a composeinfo or treeinfo file requires either to import it from a http or ftp location.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Read the doc
2. The example shows importing from http, ftp and nfs
3. Don't follow the example
If you import just from nfs then it will not work
To have it made clear that if you do this it will not work, and why.
The doc should have examples as well.
It already does have examples of what to do. The problem is that it doesn't spell out what variations of the example is valid.
(In reply to Raymond Mancy from comment #2)
> It already does have examples of what to do. The problem is that it doesn't
> spell out what variations of the example is valid.
Right, that's what I meant. Show the combinations that work, and those that don't and what gets imported, etc.
the doc page: http://beaker-project.org/docs/admin-guide/distro-import.html
Perhaps we should add a section in the architecture guide about the distro model, and what it means for a distro to be "available" in a lab?
Bug 1043318 also notes some additional constraints around the fact that we assume "available over HTTP" implies "available over NFS", and that constrains the way people configure the HTTP server providing the trees.
In particular, we need to make it clear that provisioning over a WAN is a *really* bad idea, since it wastes both time and bandwidth (unless you have a high capacity WAN optimiser that can handle implicit caching of entire operating system trees).