Suvayu Ali <fatkasuvayu+linux(a)gmail.com> wrote:
> In practice, there's not much of a difference between
> or just "clean metadata". Because both require the update/upgrade
> command to download all stuff from the network and build to whole
> meta database from scratch, even if that wouldn't be necessary.
Sorry, that's not correct. Ever since the yum days, they have been
different. The key differences are these two bits (from `man dnf'):
Sorry for any confusion I may have caused by my statement.
Of course, both are not the same, but if a user of a standard
Fedora system with no fancy configuration just wants to install
latest updates (and rarely uses dnf for anything else), then
my experience shows that there's no significant amount of cached
data that "clean metadata" would have preserved over "clean all".
No matter if "dnf clean metadata" or "dnf clean all" is used,
the directly following "dnf upgrade" uses the same amount of
resources (traffic + CPU).
Yes, I agree, other commands may benefit a lot if "clean metadata"
is used instead of "clean all". Personally, I rarely use dnf for
anything but updating the system. Maybe that's why I don't see
But to be honest, that wasn't my point anyway. My point is that
users should not need to use either of those (clean metadata or
clean all), but there should be a way that "dnf upgrade" (especially
if used interactively on command-line) provides what users expect
without any fiddling with cached data.
During the yum days, the most frequent complaint on this list were:
is yum so slow, I hate waiting for metadata updates, .... and so on.
And now people complain, the cached metadata is out of date (please
don't confuse this with actual bugs). There is no winning, you either
wait for the metadata to download, or you get a responsive interface.
For my individual needs, I run "dnf --disablerepo=fedora clean metadata"
before I check for updates (actually I use a two-line script for that).
The "fedora" repository uses up the most resources for downloading and
rebuilding metadata. This way I still benefit from caching but also get
the freshest updates.
For all other yum/dnf operations (which I rarely use) I'm fine with
the default caching and happily use the default repo files provided
PS: I've no relation to the dnf team, just a user. I however
instead of complaining on the list, if any discussions here lead you
to think there is a problem with dnf (as in a bug), you should file
Several tickets already exist for that issue, and the answer is
always the same: People should run "clean metadata" (or "clean all"
if they have plenty of resources) before they update.
Technically, that might be a proper solution, but that won't get
people from asking the very same question over and over again.
Maybe it would be less confusing if "--refresh" did the job
(which sounds like a cool workaround for that kind of problem)
but there's a difference between "--refresh" and "clean
Could that be caused by "metadata_expire=6h" in fedora-updates.repo?