[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [Pulp-list] Repository cloning design

Pradeep Kilambi wrote:
On 10/12/2010 03:56 AM, Sayli Karmarkar wrote:
I have added design approach for supporting repository cloning at:

Feel free to add your thoughts / suggestions to wiki or in this email thread.


* A cloned repo should just be another repo. Only use I see to adding a parent sub collection to a repo is to track back the parent for the clone, which we could with feed I think. But I'm open to having a parent info in the repo. But lets not call it parent_repo instead just call it a parent_id in a repo. If parent_id is None its a parent repo else its cloned.
Oh yes. What I meant is adding parent_repo_id to Repo model (sorry I was not very clear there). Like I said in reply to Todd's email, what happens when a parent repo of a feeded cloned repo is deleted? We need to either delete cloned repo or make it feedless. Also, feed does not always have parent repo id in it. We will have to go through all the repos and find one with this relative_url.

* Yes we need errata support, that you should get for free with sync, as sync should pull in updateinfo.xml and import anyway. You dont have to do anything manual here if you use a local sync.

* lets not use clone: as feed type. Just use existing 'yum' type as all repos in pulp are yum repos and point the feed to parent which is a yum repo.
+1 to that. That was an option if we don't want to add parent_id to repo and still have a way to know that this is a cloned repo.

* yea we can skip symlink support as packages are anyway symlinks in repo directory with my last sprint changes. So there is no additional duplication of rpm content.

~ Prad

Pulp-list mailing list
Pulp-list redhat com

Sayli Karmarkar
skarmark redhat com

Systems Management

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]