DNF: why does it refresh metadata all the time

Jan Zelený jzeleny at redhat.com
Fri Jun 20 07:11:06 UTC 2014

On 19. 6. 2014 at 14:24:39, Jon wrote:
> > and BTW i am not playing around that much on my Rawhide VM but had
> > *two times* today by type "dnf whatever" the "there is already an
> > instance, wating for PID..." nonsense caused by the background
> > metadata refresh
> > 
> > do you *really* think that's a good user-expierience?
> No, that is unfortunate, and probably user unfriendly.

While I agree that the user experience is not pleasant, you are talking about 
rawhide. I'd like to ask you not to use rawhide as a reference. We never 
intended to optimize dnf for the rawhide use case which is different in so many 
ways compared to stable Fedoras.

> It would be great if the metadata were fetched, and put into place
> atomically. Something where the downloading step of the fetching would
> block on YOUR pid as a user you should never lose the battle, and if
> you happen to get not a race with the finally atomic save, it would
> briefly make you wait while the stdio took place. This is one thing
> I've always though was unfortunate with yum, and would like to see
> improve with DNF. More resilient handling of tasks, or more
> concurrency, or whatever.  I guess if you were doing a clearing of
> metadata, and in the back ground metadata were already being
> fetched... the two tasks could be combined, and you really wouldn't
> need to be blocked too much.  Perhaps someday that will be
> implemented. =)

If you feel like this should be done and there is no RFE about this in 
bugzilla, feel free to file one. It will probably get deprioritized for the 
moment but it will at least be tracked.


More information about the devel mailing list