[arch-general] Why on 1st run does pacman -Qi take so long? (15 seconds waiting for Qi?)

Ralph Corderoy ralph at inputplus.co.uk
Sat Jun 2 11:28:49 UTC 2018


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


More information about the arch-general mailing list