Copr routinely swamped by nightly builds

[Valentin Gologuzov]

On 02/17/2015 01:36 PM, Kevin Fenzi wrote:

>
>> Wouldn't it be better to (c) implement some sort of 'fair' scheduling
>> instead of a pure first-come-first-serve system?
>
> Yes, that also has been talked about. I don't know where it is in
> implementation however.

I've started a feature branch to tackle this problem. At first we would 
limit number of builders allocated to the one user (important for heavy 
packages like kernel or chromium), and also re-use VM for the same user 
(big deal for coprs with bunch of small srpm). Rought ETA ~1 month.

If copr is still spawmped in future, we will introduce some dynamic 
prioritization.

-- 
Best regards,
Gologuzov Valentin.