[systemd-devel] Unit failed to load: Cannot allocate memory.

[Lennart Poettering]

On Tue, 09.08.11 07:02, Mike Kazantsev (mk.fraggod at gmail.com) wrote:

> Good day,
> 
> Updating to systemd-33 (with "include missing.h" patch) seem to give me
> the following errors on boot:
> 
>   [   27.024525] systemd[1]: Cannot add dependency job for unit fossil_echo.service, ignoring: Unit fossil_echo.service failed to load: Cannot allocate memory. See system logs and 'systemctl status fossil_echo.service' for details.
>   [   27.024912] systemd[1]: Cannot add dependency job for unit catalyst_thruk.service, ignoring: Unit catalyst_thruk.service failed to load: Cannot allocate memory. See system logs and 'systemctl status catalyst_thruk.service' for details.
>   [   27.025256] systemd[1]: Cannot add dependency job for unit uwsgi_graphite.service, ignoring: Unit uwsgi_graphite.service failed to load: Cannot allocate memory. See system logs and 'systemctl status uwsgi_graphite.service' for details.
>   [   27.025600] systemd[1]: Cannot add dependency job for unit uwsgi_ves.service, ignoring: Unit uwsgi_ves.service failed to load: Cannot allocate memory. See system logs and 'systemctl status uwsgi_ves.service' for details.
>   [   27.025989] systemd[1]: Cannot add dependency job for unit git-daemon.service, ignoring: Unit git-daemon.service failed to load: Cannot allocate memory. See system logs and 'systemctl status git-daemon.service' for details.
>   [   27.026344] systemd[1]: Cannot add dependency job for unit transmission-daemon.service, ignoring: Unit transmission-daemon.service failed to load: Cannot allocate memory. See system logs and 'systemctl status transmission-daemon.service' for details.
>   [   27.026873] systemd[1]: Cannot add dependency job for unit dnsfilter.service, ignoring: Unit dnsfilter.service failed to load: Cannot allocate memory. See system logs and 'systemctl status dnsfilter.service' for details.
> 
> Tried to reboot twice this morning (due to unrelated issues), and both
> times it were the same same errors for the same units.
> 
> systemctl status doesn't seem to show much more information:
> 
>   uwsgi_graphite.service - uwsgi: graphite web
>    Loaded: error (Reason: Cannot allocate memory)
>    Active: inactive (dead)
> 
> Rolling back to systemd-32 (w/o any other changes in the system) fixes
> the problem.

Any chance you can bisect this?

ENOMEM in these cases is very strange. It should normally not
happen. It's a bit hard to see know from this output however which
operation precisely generated the ENOMEM. 

Could you try systemd.log_level=debug systemd.log_target=kmsg and see if
you can reproduce this?

Lennart

-- 
Lennart Poettering - Red Hat, Inc.