[systemd-devel] Grouping services in systemd..

[nitish nagesh]

Hi,

 In fact i did try a similar approach of assigning CPUShares to a slice.
Basically i separated these critical services into a new slice & assigned a
CPUShare=8192. However with this i see it takes more time than before to
complete the boot. One related observation was, on my system the default
value of CPUShares was not 1024 (as mentioned in the man-pages). Instead
its a large number.






*# systemctl show system-netns.slice | grep -i
cpuCPUUsageNSec=192379569CPUAccounting=yesCPUShares=18446744073709551615StartupCPUShares=18446744073709551615CPUQuotaPerSecUSec=infinity
 *

So by setting it to 8192 am I reducing the CPUShare and hence seeing an
increase in time? I tried the same with StartupCPUShares but observations
were similar. The DefaultCPUAccounting also seems to be enabled for the
system.

Am i doing something incorrectly here? Please help.

This is the systemd version is use:
# systemctl --version
systemd 229
+PAM -AUDIT -SELINUX +IMA -APPARMOR +SMACK +SYSVINIT +UTMP -LIBCRYPTSETUP
-GCRYPT +GNUTLS +ACL +XZ -LZ4 -SECCOMP +BLKID -ELFUTILS +KMOD -IDN

Thanks,

On Thu, Apr 2, 2020 at 8:00 PM Lennart Poettering <lennart at poettering.net>
wrote:

> On Do, 02.04.20 18:51, nitish nagesh (nagesh.nitish at gmail.com) wrote:
>
> > Hi folks,
> >
> >   We are working on an embedded ARM Cortex A9 based system (aka low CPU).
> > It runs on a custom linux based operating system which uses systemd.
> >
> >   We have a bunch of daemons (around ~50+) that come up during boot
> > simultaneously which slows down the boot significantly as the CPU runs
> out
> > of breath. We were thinking of staggering these daemons into 2 groups.
> The
> > first group containing "critical" daemons (around 15) so that they finish
> > faster and make the system usable sooner. Followed by the second group of
> > daemons.
> >
> >    Separating daemons into buckets could be done using:
> > 1) systemd targets: Introduce 2 new targets and classify the
> > services/daemons into them. Layer these targets during boot.
> > 2) Cgroups: Create a new systemd slice and put all the "critical"
> services
> > into it. Allocate sufficient CPUShares value to the slice so that this
> > slice gets its due CPU% to finish faster boot.
> >
> > Can you please suggest which of the above is a better approach?
> Respective
> > pros/cons with each.
> >
> > Or if there is a third approach better than the above?
> >
> > Thanks in advance,
>
> StartupCPUWeight= and StartupIOWeight= are probably what you should be
> using? Have you played around with that? That passes this problem on
> to the CPU/IO schedulers of the kernels. i.e. you still enqueue
> everything in parallel, but tell the kernel what to schedule first.
>
> In recent systemd versions the weights configured this way also affect
> the order in which jobs are dispatched by systemd itself if multiple
> are runnable at the same time.
>
> Lennart
>
> --
> Lennart Poettering, Berlin
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/systemd-devel/attachments/20200407/f1ad9021/attachment.htm>