[f20-branch/rhel7-branch] A bit of crazy magic for caching base repo

Brian C. Lane bcl at redhat.com
Tue Oct 22 15:23:04 UTC 2013


On Tue, Oct 22, 2013 at 09:37:37AM +0200, Vratislav Podzimek wrote:
> On Mon, 2013-10-21 at 13:22 -0700, Brian C. Lane wrote:
> > On Wed, Oct 16, 2013 at 03:32:17PM +0200, Vratislav Podzimek wrote:
> > > Getting base repo from yumpayload is potentially a very expensive operation (it
> > > needs _yum_lock), but we need it quite often. To speed thinks up and prevent
> > > hangs, we may cache base repo property and invalidate the value in cases that
> > > may cause its change.
> > 
> > Oh, I missed this email the first time through. I think this should only
> > go on master for now, until we make sure there aren't side-effects.
> I forgot to include it to the patches thread. I agree it is a change
> with a high potential for side-effects and making things really bad.
> However, we should deal with the issue of text mode never starting if
> there is a kickstart with no source (repo) specified. 
> 
> Shall I port the PATCH 3/3 to rhel7-branch without using the base repo
> cache? It causes text mode to take more time to refresh in some cases,
> but I believe that's more acceptable than printing dots forever with no
> feedback on what's going on.

That sounds like a good idea to me. Just make sure it has a bz# and
acks.

-- 
Brian C. Lane | Anaconda Team | IRC: bcl #anaconda | Port Orchard, WA (PST8PDT)


More information about the anaconda-patches mailing list