Bug 33010 - The distrib-coffee.ipsl.jussieu.fr mirror is out-of-date
Summary: The distrib-coffee.ipsl.jussieu.fr mirror is out-of-date
Status: NEW
Alias: None
Product: Websites
Classification: Unclassified
Component: mirrors.mageia.org (show other bugs)
Version: trunk
Hardware: All Linux
Priority: High major
Target Milestone: ---
Assignee: Sysadmin Team
QA Contact:
URL: https://mirrors.mageia.org/status
Whiteboard:
Keywords:
: 33022 33034 (view as bug list)
Depends on: 17400
Blocks:
  Show dependency treegraph
 
Reported: 2024-03-24 13:54 CET by Frédéric "LpSolit" Buclin
Modified: 2024-04-10 15:18 CEST (History)
10 users (show)

See Also:
Source RPM:
CVE:
Status comment:


Attachments

Description Frédéric "LpSolit" Buclin 2024-03-24 13:54:39 CET
The distrib-coffee.ipsl.jussieu.fr mirror has not been synced for 5 days. Please someone contact them. This is pretty urgent as many mirrors depend on this specific mirror.
Comment 1 sturmvogel 2024-03-24 17:21:30 CET
Everybody can contact the server admin. The contact adress is directly on the front page!
Comment 2 katnatek 2024-03-25 01:30:20 CET
(In reply to sturmvogel from comment #1)
> Everybody can contact the server admin. The contact adress is directly on
> the front page!

FYI Thomas Andrews the QA leader already did it
Comment 3 sturmvogel 2024-03-27 16:17:32 CET
*** Bug 33022 has been marked as a duplicate of this bug. ***

CC: (none) => kcpammer

Comment 4 sturmvogel 2024-03-29 15:09:40 CET
*** Bug 33034 has been marked as a duplicate of this bug. ***

CC: (none) => petlaw726

Comment 5 Dave Hodgins 2024-03-29 16:18:56 CET
Blog and forum, and discuss ml posts should go out explaining that the problem.

People should be warned to check https://mirrors.mageia.org/status
for a mirror that is syncing.

Once they pick one to use, they can remove their current sources with
"urpmi.removemedia -a" (as root) and then add a specific mirror
https://wiki.mageia.org/en/Installing_and_removing_software#Adding_a_specific_Media_Mirror

CC: (none) => atelier-bugs, davidwhodgins

Dave Hodgins 2024-03-29 18:27:25 CET

CC: (none) => qa-bugs

Comment 6 PC LX 2024-03-29 21:05:52 CET
It would be good if the various Mageia tools that update packages (e.g. mgaapplet, rpmdrake) were able to detect if a repository is out of date.

Maybe the tools could look for a TIMESTAMP file in the repositories that is updated daily even if there are no new package updates, and warn the user if the repository is too out-of-date.

The warning message could suggest using another up-to-date mirror, possibly with some usability features like showing a list of up-to-date mirrors and a button to use the selected mirrors.

CC: (none) => mageia

Comment 7 Morgan Leijström 2024-03-30 12:43:25 CET
This is a too frequent problem for this important functionality.  More than once a year more than one mirror have some problem.

There was a proposal Bug 27623 - Fallback mirrors

CC: (none) => fri

Comment 8 peter lawford 2024-03-30 15:31:02 CET
since the mirror of free seems to resist, a temporarily solution is to add to /etc/yum.repos.d/mageia-x86_64.repo the uncommented line:
baseurl=http://ftp.free.fr/mirrors/mageia.org/distrib/9/x86_64/media/core/<release, updates>/ and the same for nonfree, tainted, i586 ... where the line above must be adapted to the different cases
it works well as long as the free mirror will be functional
Comment 9 peter lawford 2024-03-30 15:48:30 CET
compare to ubuntu mirrors: https://launchpad.net/ubuntu/+archivemirrors
Comment 10 Filip Komar 2024-03-30 23:17:20 CET
There are a lot of open bugs on that topic:
https://bugs.mageia.org/buglist.cgi?quicksearch=mirror

We need more permanent solution. Like the one in bug 17400 or similar.

CC: (none) => filip.komar
Depends on: (none) => 17400

Comment 11 Marc Krämer 2024-04-02 23:40:58 CEST
we have to many mirrors using distrib-coffee.ipsl.jussieu.fr as master (and it is not the first time it failed)
how can we prevent mirrors syncing from out of date mirrors or switching to a working one? can mirrorbrain provide a sync link for a protocol too, so it chooses an update mirror for sync?

btw ftp-stud.hs-esslingen.de uses "distrib-coffee.ipsl.jussieu.fr" too as a source - the status page is not correct.

Thankfully we were not hit by xz-issue, as many would not get our updates at the moment.

CC: (none) => mageia

Comment 12 katnatek 2024-04-03 00:42:28 CEST
I wonder if we not have again an issue like the ibiblio.org https://bugs.mageia.org/show_bug.cgi?id=30089#c28 ?
Comment 13 peter lawford 2024-04-03 16:30:41 CEST
I don't understand: some mirrors, as https://mageia.c3sl.ufpr.br/ and some others, are marked as being out of use since 2024-04-02 at 9:28 p.m.; it's impossible, since today, april, 3rd at 1:00 p.m. Paris time, they were green
thank for explanation
Comment 14 Dave Hodgins 2024-04-03 17:22:09 CEST
(In reply to peter lawford from comment #13)
> I don't understand: some mirrors, as https://mageia.c3sl.ufpr.br/ and some
> others, are marked as being out of use since 2024-04-02 at 9:28 p.m.; it's
> impossible, since today, april, 3rd at 1:00 p.m. Paris time, they were green
> thank for explanation

The status legend is somewhat misleading.

When a mirror shows "Less than 2 days old", it does NOT mean it's been 12 to
48 hours since it last synced.

The status is based on the age of the last package that was synced, not when
the last check for updates occurred.

If a repo has no updates for 12 hours, and then an update is built, then
all mirrors immediately become orange in status as the date/time of the hdlist
and package files that it has are at least 12 hours old. It will turn green
again as soon as the mirror syncs that repo again.

It works this way because the status page uses the date/time of the files
on the mirrors. It does NOT check the log on the main server to see when
a mirror last checked for updates. Using the log on the main server would
only be possible for the tier 1 mirrors. As the logs on the tier 1 mirrors
are not accessible, it can not be done that way for other mirrors.

For simplicity, the status of all mirrors is based on the age of the hdlist
files on the mirror.
Comment 15 peter lawford 2024-04-03 22:05:45 CEST
(In reply to Marc Krämer from comment #11)
> we have to many mirrors using distrib-coffee.ipsl.jussieu.fr as master (and
> it is not the first time it failed)
> how can we prevent mirrors syncing from out of date mirrors or switching to
> a working one? can mirrorbrain provide a sync link for a protocol too, so it
> chooses an update mirror for sync?
> 
> btw ftp-stud.hs-esslingen.de uses "distrib-coffee.ipsl.jussieu.fr" too as a
> source - the status page is not correct.
> 
> Thankfully we were not hit by xz-issue, as many would not get our updates at
> the moment.

what is xz-issue?
Comment 16 katnatek 2024-04-03 22:16:03 CEST
(In reply to peter lawford from comment #15)
> (In reply to Marc Krämer from comment #11)
> > we have to many mirrors using distrib-coffee.ipsl.jussieu.fr as master (and
> > it is not the first time it failed)
> > how can we prevent mirrors syncing from out of date mirrors or switching to
> > a working one? can mirrorbrain provide a sync link for a protocol too, so it
> > chooses an update mirror for sync?
> > 
> > btw ftp-stud.hs-esslingen.de uses "distrib-coffee.ipsl.jussieu.fr" too as a
> > source - the status page is not correct.
> > 
> > Thankfully we were not hit by xz-issue, as many would not get our updates at
> > the moment.
> 
> what is xz-issue?

https://lwn.net/Articles/967180/
But mageia is not affected https://blog.mageia.org/en/2024/03/30/about-backdoor-security-alert-for-xz/
Comment 17 Marc Krämer 2024-04-04 21:57:44 CEST
it looks like the server started syncing again.
https://mirrors.mageia.org/status

it is not up-to-date yet
Comment 18 peter lawford 2024-04-09 19:32:43 CEST
it seems that coffee-jussieu.ipsl out of synch again
Comment 19 peter lawford 2024-04-10 13:29:35 CEST
where is gone the distrib-coffee.ipsl.jussieu.fr mirror? it no loger appear in the list of mageia's mirrors
Comment 20 peter lawford 2024-04-10 15:18:29 CEST
distrib-coffee.ipsl.jussieu.fr is back and green

Note You need to log in before you can comment on or make changes to this bug.