[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: yum 3.2.23 more than 10x slower than 3.2.21?!



Joe Nall wrote:
> 
> On Jun 26, 2009, at 10:25 AM, Roberto Ragusa wrote:
> 
>> Seth Vidal wrote:
>>>
>>>
>>> On Fri, 26 Jun 2009, Roberto Ragusa wrote:
>>>
>>>> Seth Vidal wrote:
>>>>>
>>>>> and if you can make this happen every time with 3.2.23 if you could
>>>>> please run:
>>>>>
>>>>> time yum -d 3 list updates | grep 'time:'
>>>>
>>>> You were not talking to me, but I tried and
>>>> look at this: (added -C to avoid net access)
>>>>
>>>> # time yum -C -d 3 list updates --disablerepo=atrpms\* | grep 'time:'
>>>> Config time: 0.271
>>>> repo time: 0.001
>>>> pkgsack time: 86.684
>>>
>>> and you're POSITIVE nothing is being downloaded here? Nothing at all?
>>>
>>> Run it again, please and capture all the output and post it to a
>>> pastebin.
>>>
>>
>> I have a hint and it points to the "same rpms in more than one repo"
>> hypothesis.
>>
>> yum -d 5 lists a lot of
>>
>> excluding for cost: firefox-3.0.11-1.fc10.i386 from updates
>> excluding for cost: gnome-session-2.24.3-1.fc10.i386 from updates
>> [...]
>>
>> I see something related to cost handling fixes in the yum Changelog.
>>
>> My guess is that elements are removed from an array one at a time
>> and everything after that is moved back one position.
>> Or some inefficiency of that kind.
>>
>> You may be able to reproduce it by configuring two updates repos.
> 
> The performance change occurred between 3.2.21-2 and 3.2.22-5 based on
> testing on koji f10 downloads.

from rpm -q -changelog yum:

* Tue May 12 2009 Seth Vidal <skvidal at fedoraproject.org> - 3.2.22-5


from ChangeLog in yum tar.gz:

2009-04-13  James Antill <james and org>

	* yum/yumRepo.py: Fix cost sorting of repos.


-- 
   Roberto Ragusa    mail at robertoragusa.it


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]