You are not logged in.
Running Arch on my ThinkPad, when I -Syu, I usually get something like this:
$ time pacman -Syu
:: Synchronizing package databases...
core is up to date
extra is up to date
community is up to date
:: Starting full system upgrade...
there is nothing to do
real 1m29.062s
user 0m0.560s
sys 0m0.610s
It gets to Starting full system upgrade within a few (3 or 4) seconds, and then spends the rest of its time there. During that time, my hard drive activity light is going, but my drive is quite fast, being a 7200RPM Scorpio Black. When I ran Arch on an old Dell laptop (with my same home network connection and using the same mirrors), I would usually complete the same operation within 5-10 seconds.
I don't think it's a network issue -- it happens on different wireless networks that I connect to, and I've tried a number of different mirrors. When I do install a package, it usually downloads at around 200-500KB/s, so that part is actually quite fast. My /var partition is ext4.
I've done pacman -Sc and pacman-optimize, which hasn't helped. I also run Arch on my Linode VPS (and yes I realize they are using faster drives than I have), but the same operation there completes with real 0m0.494s.
Anybody have any idea why it's this slow?
Last edited by MkFly (2010-11-14 18:23:25)
Offline
Have you tried
pacman -Syy
to see if that fixes that problem ?
Offline
Have you tried
pacman -Syy
to see if that fixes that problem ?
Hadn't thought of that -- just gave it a try, didn't help.
Something I noticed: on a fresh boot, it takes this long amount of time. After that, it's as quick as it should be (2-5 seconds), unless I wait a while (say, more than 10 minutes) and try it again, in which case it's slooooww again. Also, this doesn't just happen with -Syu, it's slow like this for pretty much everything (-Ss, -Qs, -Ql, -Qo, -Qi, etc.)
Offline
Offline
Thanks. pacman-cage is a BIG help:
$ time pacman -Syu
:: Synchronizing package databases...
core is up to date
extra is up to date
community is up to date
:: Starting full system upgrade...
there is nothing to do
real 0m4.142s
user 0m0.323s
sys 0m0.360s
So, is the underlying cause that the all of those files in the database are stored non-contiguously?
Last edited by MkFly (2010-11-14 08:37:07)
Offline
Yes... with pacman-3.4, only new database entries from the sync db were extracted. That was a big speed up, but caused the sync database to be spread across the filesystem, so caused long term slow down.
With pacman-3.5, we now read the sync dbs directly from the tarballs that are downloaded so no fragmentation there. Also, the local database may get somewhat more fused, with the "depends" and "desc" files being merged together. Overall, pacman-3.5 will show a very noticeable speed up.
Online
Yes... with pacman-3.4, only new database entries from the sync db were extracted. That was a big speed up, but caused the sync database to be spread across the filesystem, so caused long term slow down.
With pacman-3.5, we now read the sync dbs directly from the tarballs that are downloaded so no fragmentation there. Also, the local database may get somewhat more fused, with the "depends" and "desc" files being merged together. Overall, pacman-3.5 will show a very noticeable speed up.
Thanks Allan.
Offline
Wouldnt pacman-optimize help too?
Offline
The idea behind pacman-optimize is to move all your database files to one continuous block on the hard-drive. That is the theory... in practice there may be a different story.
Online