On 02/26/2015 06:14 PM, Kevin Fenzi wrote:
On Thu, 26 Feb 2015 17:22:01 +0100
Ralf Corsepius <rc040203@xxxxxxxxxx> wrote:
And would you have the kindness to tell me what I can to about it?
I'm not sure. It's some issue with your communication to the koji
hub...
Meanwhile I rebooted (for unrelated reasons), now the ssl-errors are
gone. So, it have been could be your or my end - No idea.
Can you do a:
koji list-tasks --mine
# koji list-tasks --mine
Error: [('SSL routines', 'SSL3_READ_BYTES', 'ssl handshake failure')]
Is your fedora koji cert up to date?
Yes it is.
Do:
fedora-cert -v
and if it's expired it should offer to issue you a new one.
Usually however, it would say expired, not handshake failure.
Can you ping koji.fedoraproject.org? browse to it?
Is the time correct on your machine?
Yes, yes and yes ... BTW: koji/fedpkg now seems to work again.
I am working on tracking down the problem, but it's proving quite
elusive. ;(
BTW: Here's another variant of a break down:
https://kojipkgs.fedoraproject.org//work/tasks/1835/9081835/root.log
This time, yum/dnf (whatever currently is being used, I presume
it's yum) is demonstrating one of the yum/dnf issues we discussed
yesterday:
Yum retries to download packages from a mirror it could have know
to be broken, because it failed to connect to it before.
Sure, but then it would have just failed faster as thats the only
mirror thats defined internally for builders.
yum could have tried a different mirror instead of retrying the
already broken one again.
There's no other mirrors for internal builds. It's a baseurl to the
kojipkgs server.
Yes, my point is it is X times trying in vain to access a host to
download X packages. Not sure what happens when this happens with a real
mirrorlist/metalink-list, whose top 5 are unreachable/flakey/dead or broken.
I am sure I've seen, yum iterating X times over the same list of broken
mirrors.
Ralf
--
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxxx
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
