On Fri, 8 Feb 2019 at 16:00, Stephen John Smoogen <smooge(a)gmail.com> wrote:
On Fri, 8 Feb 2019 at 14:31, Adam Williamson <adamwill(a)fedoraproject.org> wrote:
>
> On Thu, 2019-02-07 at 12:21 -0500, Stephen John Smoogen wrote:
> > On Thu, 7 Feb 2019 at 09:40, Stephen John Smoogen <smooge(a)gmail.com>
wrote:
> > > On Thu, 7 Feb 2019 at 09:21, Fabio Valentini <decathorpe(a)gmail.com>
wrote:
> > > > On Thu, Feb 7, 2019 at 3:01 PM Stephen John Smoogen
<smooge(a)gmail.com> wrote:
> > > > > On Thu, 7 Feb 2019 at 07:34, Fabio Valentini
<decathorpe(a)gmail.com> wrote:
> > > librepo.LibrepoException: (8, 'Cannot prepare internal mirrorlist:
> > > Curl error (28): Timeout was reached for
> > >
https://mirrors.fedoraproject.org/metalink?repo=updates-released-f29&...
> > > [Operation too slow. Less than 1000 bytes/sec transferred the last 30
> > > seconds]', 'An Curl handle error')
> > >
> > > That says the proxy it was trying to get to was too slow.. I will see
> > > if I can find which one it was at that timezone and see if I can
> > > address.
> > >
> >
> > OK the problem does not look like it is mirrorlist related but
> > something to do with the copr host/network at that time
> >
> > 209.132.184.33 - - [07/Feb/2019:00:01:59 +0000] "GET
> > /metalink?repo=updates-released-f29&arch=x86_64 HTTP/1.1" 200 18343
> > "-" "dnf/2.7.5"
> >
> > says it sent the data but it looks like something that the client was
> > looking for timed out. I will work with the copr team to diagnose
> > further.
>
> FWIW, openQA tests have been failing relatively frequently with
> timeouts while refreshing repos or 503s from mirrormanager for the last
> few weeks. I'm pretty sure this didn't used to be anything like as
> common...
I am checking if we are seeing an increase and when. There are
generally some every hour when mirrormanager changes out pkls but
sometimes it grows.
OK looking at the general stats for the last 2 months, we were seeing
an average of 0.2% of all connections generating a 503 (Out of
22114356 connections on 2018/12/03 only 41263 were 503 and the rest
are 200, similar to all days). There doesn't seem to be any large
uptick in them that stands out but I am not ruling out specific
mirrors, times of requests or clients yet that could make these more
likely.
--
Stephen J Smoogen.