#6366: missing metadata about nightly rawhide/24 composes
------------------------------+-----------------------
Reporter: bpeck | Owner: rel-eng@…
Type: defect | Status: new
Milestone: Fedora 24 Alpha | Component: other
Resolution: | Keywords:
Blocked By: | Blocking:
------------------------------+-----------------------
Comment (by ausil):
I am honestly not sure the best way to resolve this. A big part of the
problem is taht we split the compose up after it is done.
Lets look at a nigtly f24 compose, The compose is done in
https://kojipkgs.fedoraproject.org/compose/branched/Fedora-24-20160318.n....
We have a bunch of directories that have the various parts of the compose
{{{
Atomic Cloud CloudImages Docker Everything Labs metadata Server
Spins Workstation
}}}
the metadata directory has a composeinfo.json file in it that describes
nicely all the content of the compose
however, when we put the compose on the mirrors, we go and carve it up.
{{{
Atomic CloudImages Docker Everything Server Spins Workstation
}}}
all go into the main public location
http://dl.fedoraproject.org/pub/fedora/linux/development/24/
we then put the remianing deliverables over in alt
{{{
Cloud Labs
}}}
http://dl.fedoraproject.org/pub/alt/development/24/
throwing metadata on the floor as it now is incorrect.
there is .treeinfo files all through the compose but you need to tell
beaker where to look for them.
we have a ticket
https://pagure.io/pungi/issue/98 open to enable us to
change things up to better support our use cases.
AFAIK the only use case for .composeinfo is beaker. dmach would be the
person that made the change from .composeinfo to composeinfo.json
another way to solve this is pungi gives a message when it is done
composing, it is follow up with a message later that says the rsync is
done. If you use the message from pungi you could fetch the
composeinfo.json and prep things for when the rsync is done. knowing how
we carve things up. It is not as ideal as it could be as we would need to
let you know if we change how we carve things up, but I do not anticipate
changing that before we get pungi being able to better support our use
cases.
Secondary arches are a whole different problem. Since we are merging three
composes into one.
--
Ticket URL: <
https://fedorahosted.org/rel-eng/ticket/6366#comment:1>
Fedora Release Engineering <
http://fedorahosted.org/rel-eng>
Release Engineering for the Fedora Project