[arch-general] Why on 1st run does pacman -Qi take so long? (15 seconds waiting for Qi?)
This is new, after the latest pacman update along with 4.6.13, when running
$ pacman -Qi somepackage
for the first time it takes 15 seconds for any response. E.g.,
$ time pmqi colord
Name : colord
Version : 1.4.3-1
Description : System daemon for managing color devices
Architecture : x86_64
URL : https://www.freedesktop.org/software/colord
Licenses : GPL2
Groups : None
Provides : None
Depends On : lcms2 libgusb polkit sqlite systemd dconf dbus
libgudev shared-mime-info
Optional Deps : argyllcms: color profiling
colord-sane: SANE support
Required By : gtk3
Optional For : cups
Conflicts With : None
Replaces : shared-color-profiles
Installed Size : 6.60 MiB
Packager : Jan Alexander Steffens (heftig)
Hi David,
This is new, after the latest pacman update along with 4.6.13, when running $ pacman -Qi somepackage for the first time it takes 15 seconds for any response.
I've always found it sluggish if its data isn't cached in RAM. `vmstat 1' shows one core is pegged out waiting for I/O; the `wa' column. The I/O is the reading of 814 /var/lib/pacman/local/*/desc. $ LC_ALL=C strace -c pacman -Qi ed % time seconds usecs/call calls errors syscall ------ ----------- ----------- --------- --------- ---------------- 50.40 0.157668 185 853 openat 35.51 0.111096 67 1670 read 6.27 0.019606 23 853 close 5.48 0.017140 20 856 fstat 0.76 0.002383 34 70 mmap First run of `\time -v pacman -Qi ed' v. second, boring lines deleted. Command being timed "pacman -Qi ed" "pacman -Qi ed" User time (seconds) 0.10 0.06 System time (seconds) 0.29 0.02 Percent of CPU this job got 2% 96% Elapsed (wall clock) time 0:14.70 0:00.08 Voluntary context switches 1786 0 Involuntary context switches 26 16 File system inputs 34768 0 File system outputs 0 0 -- Cheers, Ralph. https://plus.google.com/+RalphCorderoy
participants (2)
-
David C. Rankin
-
Ralph Corderoy