pacman -Ss / yaourt -Ss : sometimes *very* slow

stefan1975 · 2009-08-17 14:06:11

Hi,

i have noticed that sometimes - but not always for some reason - searching with pacman (3.3) or yaourt is Really slow. Sometimes it takes a minute or longer before it starts showing results, but when they appear they scroll by fast enough. In between running pacman and getting the results I see the HDD indicator going berserk.

The next few searches after this horrendous slow one give instant response, it seems like it cached something at that time or something, but the delay is *much* longer then a [repo] refresh.

I am running a fresh install on x86_64 with pacman 3.3 on a 7200rpm HDD and ext4.
so what would be causing this and how can i speed this up? I use pacman-color 3.3 from AUR but disabling that did not help either.

stefan

flamelab · 2009-08-17 15:42:58

Maybe, you should change your mirror ? The french mir and mir2 are fast

stefan1975 · 2009-08-17 15:47:33

i really do not think it has something to with the mirrors, i have done a Syu often just prior to searching and tried the archlinux.fr and the cs.vt.edu mirrors which are pretty snappy for me, plus the delay seems to be related to the HDD going berserk at one search and then not for some time after that ?!

for example a "yaourt -Ss networkmanager" (to include AUR for GIT versions) took almost two minutes on my pretty decent hardware.

wain · 2009-08-17 20:17:39

Run pacman-optimize or just use pacman-cage

Last edited by wain (2009-08-17 20:18:15)

stefan1975 · 2009-08-17 20:31:34

i tried pacman-optimize almost once every day. i heard some dubious opinions about caching pacman into memory, so it would not be first on my list of things to try. besides it is not always slow, when it runs it runs fine ... just once every few searches it must be trying to do/cache/check/verify/refresh or something that makes me wait for minutes on end.

thanks for the pointers thought.
s.

deej · 2009-08-17 20:36:41

A lot of people are reporting this - I also noted a slowdown.
Somewhere, in another thread, someone suggested running 'pacman -Syy' to
'clear the blockage', and this seems to have worked - for me, at least.

Maybe worth a go...

Deej

shining · 2009-08-18 09:01:02

stefan1975 wrote:

Hi,
i have noticed that sometimes - but not always for some reason - searching with pacman (3.3) or yaourt is Really slow. Sometimes it takes a minute or longer before it starts showing results, but when they appear they scroll by fast enough. In between running pacman and getting the results I see the HDD indicator going berserk.
The next few searches after this horrendous slow one give instant response, it seems like it cached something at that time or something, but the delay is *much* longer then a [repo] refresh.
I am running a fresh install on x86_64 with pacman 3.3 on a 7200rpm HDD and ext4.
so what would be causing this and how can i speed this up? I use pacman-color 3.3 from AUR but disabling that did not help either.
stefan

Maybe a hard disk problem?
Did you check dmesg?

stefan1975 · 2009-08-18 10:43:47

i am quite confident that it is no hardware problem. there are no issues reported in the logfiles. I also run windows7 and fedora on this laptop without issues. From reading other threads it seems like it is more of a widespread problem, more people seem to be experiencing pacman issues since 3.3. I have also noticed that doing a pacman -Syy indeed seems to help, or at least in the cases I have tried it since yesterday.

shining · 2009-08-18 12:05:39

stefan1975 wrote:

i am quite confident that it is no hardware problem. there are no issues reported in the logfiles. I also run windows7 and fedora on this laptop without issues. From reading other threads it seems like it is more of a widespread problem, more people seem to be experiencing pacman issues since 3.3. I have also noticed that doing a pacman -Syy indeed seems to help, or at least in the cases I have tried it since yesterday.

I didn't see any other references of -Ss being slow.

stefan1975 · 2009-08-18 12:46:51

well perhaps not -Ss perse, but i somehow feel it is related to this thread:

http://bbs.archlinux.org/viewtopic.php?id=77766

perhaps i have not investigated intesively enough yet and the problem occurs with -Syu as well. I do think as i mentioned that it has something to do with the disk (cache) and the repos and especially when the local repos are up to date already, so far this problem has not occures with a forced refresh of the mirrors yet (although after a day and a half that still might be coincidence).

stefan

shining · 2009-08-18 13:01:21

stefan1975 wrote:

well perhaps not -Ss perse, but i somehow feel it is related to this thread:
http://bbs.archlinux.org/viewtopic.php?id=77766
perhaps i have not investigated intesively enough yet and the problem occurs with -Syu as well. I do think as i mentioned that it has something to do with the disk (cache) and the repos and especially when the local repos are up to date already, so far this problem has not occures with a forced refresh of the mirrors yet (although after a day and a half that still might be coincidence).
stefan

the issues are completely different ...

-Syy is much more disk intensive than -Sy.
The only reason -Sy can be slower is because of a long network timeout. But it does not use the disk at all... It's a network problem.

However -Ss always scan the whole database, so if it's slow, it is a disk / filesystem problem.
And -Syy can indeed have an impact here, because it recreates the databases.

Arch Linux

#1 2009-08-17 14:06:11