[systemd-devel] Fwd: Journalctl performance

[Sébastien Luttringer]

On Wed, Oct 2, 2013 at 12:31 AM, David Strauss <david at davidstrauss.net> wrote:
> On Tue, Oct 1, 2013 at 6:13 AM, Colin Guthrie <gmane at colin.guthr.ie> wrote:
>> Ouch 5s for a status is nasty.
>
> We regularly see this on our production systems. Yes, it's unfortunate.

Yes, we are probably a lot to suffer of this. Today, 2mn for a status,
1m10s for the last line on my main personal server.

# echo 3 > /proc/sys/vm/drop_caches
# time systemctl status postfix.service
postfix.service - Postfix Mail Transport Agent
   Loaded: loaded (/usr/lib/systemd/system/postfix.service; enabled)
   Active: active (running) since sam. 2013-10-12 20:29:22 CEST; 6h ago
  Process: 248 ExecStart=/usr/bin/postfix start (code=exited, status=0/SUCCESS)
 Main PID: 522 (master)
   CGroup: /system.slice/postfix.service
           ├─  522 /usr/lib/postfix/master -w
           ├─  524 ...
oct. 13 03:02:07 ...
Hint: Some lines were ellipsized, use -l to show in full.
systemctl status postfix.service  0,01s user 0,55s system 0% cpu 1:53,83 total

and 1mn11s for the last line:

# time journalctl -a -n 1
-- Logs begin at sam. 2013-06-08 19:30:26 CEST, end at dim. 2013-10-13
03:13:01 CEST. --
oct. 13 03:13:01 achille.seblu.net fcron[31606]: Job php
/usr/share/webapps/owncloud/cron.php completed
journalctl -a -n 1  0,00s user 0,41s system 0% cpu 1:10,80 total

Time is constant for multiple tries with a drop_caches before.
The box is a dual core intel core i3 3.10Ghz, 4G DDR3 and 2 7200RPM
drives with btrfs and archlinux. Drive linear throughput is 130MB/s.
It is becoming painful to query the journal...

Cheers,

-- 
Sébastien "Seblu" Luttringer
https://www.seblu.net
GPG: 0x2072D77A