On 2/11/19 5:46 AM, Nicolas Mailhot wrote:
Le 2019-02-11 13:51, Fabio Valentini a écrit :
> Either I am unreasonably unlucky, or your statistics are wrong:
> Looking at [0], 35 of the last 44 build tasks failed due to network
> issues (not counting all sub-tasks), which is about 80%.
You're not unreasonably unlucky, and the stats are not wrong. You're
just not talking about the same things.
A repo sync that works will results in hundreds or more of requests just
to get the various packages and files. So it will easily account for
hundreds or more of successful requests in stats. OTOH a failing sync
will account for at most 4-5 retries before dnf gives up. So the raw
stat number is heavily skewed against counting actual copr build failures.
This is made worse by:
https://bugzilla.redhat.com/show_bug.cgi?id=1643281
basically dnf gets a mirror, downloads packages from it, but if it gets
a 503, it just retries the same mirror and gives up, it doesn't go to
the next mirror. ;(
kevin