Persistently out of date mirror

[Thomas Klar]

Hello Mirror Admins,

I've just signed up to report this issue, which I've discovered last 
week, that one of the two mirrors in Switzerland is out of date and has 
been so for months. CERN has finally fixed their mirror, so it doesn't 
hurt that much any more, but the Init7 mirror has not been updated since 
November 2025:

https://mirror.init7.net/fedora/fedora/linux/updates/43/Everything/x86_64/Pa...

I've contacted them directly, asking them to either resume syncing or 
shut down their mirror, but they wouldn't even respond. Please remove 
them from the list, because for IPs from Switzerland, CERN and Init7 are 
the first 2 on the list being generated. If Init7 ends up being the 
first server contacted, updates are rendered inaccessible by their 
mirror being live, but out of date.

Thank you in advance,

Thomas

[Adrian Reber]

On Sat, Feb 14, 2026 at 01:08:18PM +0100, Thomas Klar wrote:
> I've just signed up to report this issue, which I've discovered last week,
> that one of the two mirrors in Switzerland is out of date and has been so
> for months. CERN has finally fixed their mirror, so it doesn't hurt that
> much any more, but the Init7 mirror has not been updated since November
> 2025:
>
> https://mirror.init7.net/fedora/fedora/linux/updates/43/Everything/x86_64/Pa...
>
> I've contacted them directly, asking them to either resume syncing or shut
> down their mirror, but they wouldn't even respond. Please remove them from
> the list, because for IPs from Switzerland, CERN and Init7 are the first 2
> on the list being generated. If Init7 ends up being the first server
> contacted, updates are rendered inaccessible by their mirror being live, but
> out of date.

How are getting the information to use that mirror. The only reliable
interface for mirrors is the metalink? Are you using metalinks or
something else?

Adrian

[Thomas Klar]

Le 14.02.26 à 21:35, Adrian Reber via Mirror-admin a écrit :
> On Sat, Feb 14, 2026 at 01:08:18PM +0100, Thomas Klar wrote:
> > I've just signed up to report this issue, which I've discovered last week,
> > that one of the two mirrors in Switzerland is out of date and has been so
> > for months. CERN has finally fixed their mirror, so it doesn't hurt that
> > much any more, but the Init7 mirror has not been updated since November
> > 2025:
> >
> > https://mirror.init7.net/fedora/fedora/linux/updates/43/Everything/x86_64/Pa...
> >
> > I've contacted them directly, asking them to either resume syncing or shut
> > down their mirror, but they wouldn't even respond. Please remove them from
> > the list, because for IPs from Switzerland, CERN and Init7 are the first 2
> > on the list being generated. If Init7 ends up being the first server
> > contacted, updates are rendered inaccessible by their mirror being live, but
> > out of date.
>
> How are getting the information to use that mirror. The only reliable
> interface for mirrors is the metalink? Are you using metalinks or
> something else?
> Adrian

Oh, right. I was still using the mirrorlist. I have some custom scripts 
and config files that I use with dnf reposync to create my local 
mirrors, which I then share via NFS to all my local hosts.

So, the XML of the metalink has some hashes, which then allows the 
exclusion of an outdated mirror. Is that it?

Thanks for the suggestion. I have initially just switched to using one 
specific mirror with baseurl, but the metalink is a better solution.

None the less, does this also mean, that there's no policing of mirrors 
that serve outdated or otherwise incorrect files? The mirror list has 
not yet been deprecated and there is a humanreadable page with the list 
of mirror servers, both listing this Init7 mirror. So even if my 
detection of the issue was due to the use of an outdated access method, 
I still stand by my report of this being a bad mirror, in that it is 
several months out of date.

Thomas

[Adrian Reber]

On Sat, Feb 14, 2026 at 11:24:30PM +0100, Thomas Klar wrote:
> Le 14.02.26 à 21:35, Adrian Reber via Mirror-admin a écrit :
> > On Sat, Feb 14, 2026 at 01:08:18PM +0100, Thomas Klar wrote:
> > > I've just signed up to report this issue, which I've discovered last week,
> > > that one of the two mirrors in Switzerland is out of date and has been so
> > > for months. CERN has finally fixed their mirror, so it doesn't hurt that
> > > much any more, but the Init7 mirror has not been updated since November
> > > 2025:
> > >
> > > https://mirror.init7.net/fedora/fedora/linux/updates/43/Everything/x86_64/Pa...
> > >
> > > I've contacted them directly, asking them to either resume syncing or shut
> > > down their mirror, but they wouldn't even respond. Please remove them from
> > > the list, because for IPs from Switzerland, CERN and Init7 are the first 2
> > > on the list being generated. If Init7 ends up being the first server
> > > contacted, updates are rendered inaccessible by their mirror being live, but
> > > out of date.
> >
> > How are getting the information to use that mirror. The only reliable
> > interface for mirrors is the metalink? Are you using metalinks or
> > something else?
>
> Oh, right. I was still using the mirrorlist. I have some custom scripts and
> config files that I use with dnf reposync to create my local mirrors, which
> I then share via NFS to all my local hosts.
>
> So, the XML of the metalink has some hashes, which then allows the exclusion
> of an outdated mirror. Is that it?

It is complicated. We are not able to verify the content of all mirrors
in reasonable time (<12 hours), so we give you the hash of the correct
file and, as you said, the client can use that mirror if the hash
matches or go to the next mirror until a matching hash is found.
> Thanks for the suggestion. I have initially just switched to using one
> specific mirror with baseurl, but the metalink is a better solution.
>
> None the less, does this also mean, that there's no policing of mirrors that
> serve outdated or otherwise incorrect files? The mirror list has not yet
> been deprecated and there is a humanreadable page with the list of mirror
> servers, both listing this Init7 mirror. So even if my detection of the
> issue was due to the use of an outdated access method, I still stand by my
> report of this being a bad mirror, in that it is several months out of date.

Also complicated ;-). The EPEL mirror is up to date and we can only
disable a complete mirror (we could delete the category), so it seems to
be only a problem of the Fedora category. At the same time there are
still files which can be served from that mirror for parts which have
not changed for some time and the dnf should handle it. So it is not
completely useless. I agree that it is definitely not optimal.

I disabled the mirror for now. Important is that you already contacted
them. Thanks.

Adrian