Re: [gentoo-user] problems with performance when booted using systemd

[J. Roeleveld]

On Friday, June 06, 2014 04:46:35 AM covici@××××××××××.com wrote:
> J. Roeleveld <joost@××××××××.org> wrote:
> > On Friday, June 06, 2014 03:45:17 AM covici@××××××××××.com wrote:
> > > J. Roeleveld <joost@××××××××.org> wrote:
> > > > On Friday, June 06, 2014 01:59:18 AM covici@××××××××××.com wrote:
> > > > > Hi.  I am having some strange performance problems when booted under
> > > > > systemd.  These problems happened a little bit under openrc, but are
> > > > > much more pronounced with systemd.
> > > >
> > > > I don't think it's necessarily systemd itself, just a setting that
> > > > systemd
> > > > does differently then openrc. See below for more.
> > > >
> > > > > I am using just virtual consoles, no gui whatsoever at the moment.
> > > > > I
> > > > > also use tmux with 4 windows in one of the vcs.  My system is an i7
> > > > > processor, quod core and 16g of ram and 2g of swap space which
> > > > > appears
> > > > > not to be used.  I am using uvesafb for the console, so I get 64x160
> > > > > screens.
> > > >
> > > > Sounds similar to my laptop, except I run KDE and got 16g of swap (for
> > > > hibernate)
> > > >
> > > > > The first problem is that if I don't press any keystrokes for
> > > > > several
> > > > > minutes and then want to move to another vc, it takes about 3 or 4
> > > > > seconds after the alt-left arrow or alt-right arrow command to take
> > > > > effect.  Even within the same vt, if I don't do anything for several
> > > > > minutes, it takes several seconds till the keystroke echoes and
> > > > > something happens.  Once I have done this, things act normally, but
> > > > > its
> > > > > kind of annoying.
> > > >
> > > > Sounds like a powersave setting. I used to get the same on my old
> > > > laptop
> > > > with spinning rust. SSDs tend to "spin-up" a lot quicker.
> > > >
> > > > > Also, my load average seems to always be >1.  I have
> > > > > looked at top and things seem to be OK, except that my cpu usage is
> > > > > like
> > > > > this:
> > > > > Tasks: 934 total,   2 running, 931 sleeping,   0 stopped,   1 zombie
> > > > > %Cpu(s): 12.5 us,  1.2 sy,  0.0 ni, 86.0 id,  0.2 wa,  0.0 hi,  0.0
> > > > > si,
> > > > > 0.0 st
> > > > > KiB Mem:  16450248 total,  9678656 used,  6771592 free,  1084088
> > > > > buffers
> > > > > KiB Swap:  2097148 total,        4 used,  2097144 free.  1147688
> > > > > cached
> > > > > Mem
> > > > >
> > > > >   PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+
> > > > >   COMMAND
> > > > >
> > > > >  9969 root      20   0     708     16      0 R 100.0  0.0   1549:10
> > > > >  v86d
> > > > >
> > > > >   579 root      30  10       0      0      0 S   9.1  0.0  16:09.93
> > > > >   speakup
> > > > >
> > > > > 11789 root      20   0   22524   2388   1116 R   0.7  0.0   0:00.03
> > > > > top
> > > > >
> > > > >     7 root       0 -20       0      0      0 S   0.3  0.0   0:10.41
> > > > >
> > > > >   kworker/u:0H
> > > > >
> > > > > and onward ...
> > > > > This is an awful lot of tasks, I have never seen so many!
> > > >
> > > > That is a lot, I am currently running KDE, firefox and a citrix remote
> > > > desktop thing. (oh, and skype and kopete and a few other items)
> > > > KDE is installed with semantic-desktop, but the nepomuk stuff is
> > > > disabled
> > > > in system-settings.
> > > > I have 200 tasks (yes, nice round figure)
> > > >
> > > > > Anyone have any ideas?  Thanks much.
> > > >
> > > > For the amount of tasks, check that you are not starting too many
> > > > unneeded
> > > > services. For the load-average of 1, shouldn't be too much of an
> > > > issue,
> > > > had
> > > > similar in the past with a lot of stuff running and slow disks.
> > > >
> > > > For the freezing, I would suggest checking all the powersave options,
> > > > especially the ones for the harddrives.
> > > > Is there anything in the logs when this happens? Eg. check the logs
> > > > right
> > > > after the system becomes responsible again, maybe there is a hint
> > > > there
> > > > what is causing this.
> > >
> > > Unless systemd is setting some powersave options, I certainly never set
> > > anything like that, this is a desktop  machine, not even a laptop.  Next
> > > time this happens I will check the logs.  Does systemd set some
> > > powersave options by default?
> >
> > I do not know that for sure, best wait for more knowledgable systemd users
> > to answer that. If it doesn't, then systemd itself is causing more
> > freezes (as per your experience) then openrc.
> >
> > I would guess it does or at least with the default configuration. What you
> > describe makes me think the disks are switched to powersave sooner with
> > systemd.
> > Can you provide the output of the following command:
> > #  hdparm -B /dev/sda
> > to get the APM settings of the disk. (If you have multiple disks, please
> > run it for the others as well.
> >
> > Question for others as well, how do you get the current setting for the
> > spindown timeout set with "  hdparm -S <value> <device> "?
> > I couldn't find it.
> >
> > I am happy with openrc and have no intention on switching to systemd as I
> > haven't heard of a single feature that would actually make my life easier.
>
> I don't have hdparm on the system, is it only for  older disks?  If
> memory serves, it did not work at all when I tried it as my disks are
> all /dev/sda, etc, but that may be wrong.

It also works on new SATA drives and SSDs:

# smartctl -a /dev/sda
smartctl 6.1 2013-03-16 r3800 [x86_64-linux-3.12.20-gentoo] (local build)
Copyright (C) 2002-13, Bruce Allen, Christian Franke, www.smartmontools.org

=== START OF INFORMATION SECTION ===
Device Model:     INTEL SSDMCEAC120B3
Serial Number:    CVLI3223002B120E
LU WWN Device Id: 5 5cd2e4 00028738f
Firmware Version: LLLi
User Capacity:    120,034,123,776 bytes [120 GB]
Sector Size:      512 bytes logical/physical
Rotation Rate:    Solid State Device
Device is:        Not in smartctl database [for details use: -P showall]
ATA Version is:   ACS-2 (minor revision not indicated)
SATA Version is:  SATA 3.0, 6.0 Gb/s (current: 6.0 Gb/s)
Local Time is:    Fri Jun  6 12:46:59 2014 CEST
SMART support is: Available - device has SMART capability.
SMART support is: Enabled

# hdparm /dev/sda

/dev/sda:
 multcount     = 16 (on)
 IO_support    =  1 (32-bit)
 readonly      =  0 (off)
 readahead     = 256 (on)
 geometry      = 14593/255/63, sectors = 234441648, start = 0

# hdparm -B /dev/sda

/dev/sda:
 APM_level      = 254

There might also be other ways to configure the powermanagement settings, I 
haven't looked into those yet.

--
Joost