Re: Some services cannot start at boot time because /run or /var is not mounted

[Dmitry Katsubo <dma_k@mail.ru>]

Dear Debian community,

I hit the similar problem but this time with /run folder. Few services have
failed to start:

# systemctl status php7.0-fpm.service
Jun 24 23:09:48 debian php-fpm7.0[893]: [24-Jun-2020 23:09:48] ERROR: unable to bind listening socket for address '/run/php/php7.0-fpm.sock': No such file or directory (2)
Jun 24 23:09:48 debian php-fpm7.0[893]: [24-Jun-2020 23:09:48] ERROR: FPM initialization failed
Jun 24 23:09:48 debian systemd[1]: php7.0-fpm.service: Main process exited, code=exited, status=78/CONFIG
Jun 24 23:09:48 debian systemd[1]: php7.0-fpm.service: Failed with result 'exit-code'.
Jun 24 23:09:48 debian systemd[1]: Failed to start The PHP 7.0 FastCGI Process Manager.

# systemctl status motioneye.service
Jun 24 23:09:47 debian systemd[1]: Started motionEye Server.
Jun 24 23:09:48 debian meyectl[895]:     INFO: hello! this is motionEye server 0.41
Jun 24 23:09:48 debian meyectl[895]: CRITICAL: pid directory "/run/motioneye" does not exist or is not writable
Jun 24 23:09:48 debian systemd[1]: motioneye.service: Main process exited, code=exited, status=255/EXCEPTION
Jun 24 23:09:48 debian systemd[1]: motioneye.service: Failed with result 'exit-code'.

# cat /etc/tmpfiles.d/php.conf
d /run/php/sessions 1733 root root

# cat /etc/tmpfiles.d/motioneye.conf
d /run/motioneye 0750 motion motion

Just after the boot I have inspected /run folder. It was created/mounted
correctly and there have been a lot of files/directories there. I suspect that
all services that have created the necessary directory under /run were able to
start normally. Few of them which relied on existence of specific directory,
have failed to started. After I have replayed the corresponding instructions for
tmpfiles.d, the services have started normally.

I have a feeling that systemd-tmpfiles was executed before /run was mounted.

Needless to note that the problem is not persistent: sometimes OS boots without
a single failed service.

How can I debug the problem?

Thank you!

On 2020-05-18 02:39, Dmitry Katsubo wrote:
> On 2020-05-11 20:11, Darac Marjal wrote:
>> On 11/05/2020 08:40, Reco wrote:
>>> 	Hi.
>>>
>>> On Mon, May 11, 2020 at 09:33:59AM +0200, Dmitry Katsubo wrote:
>>>
>>>> root@debian:~ # systemctl status binfmt-support
>>>> * binfmt-support.service - Enable support for additional executable binary formats
>>>>    Loaded: loaded (/lib/systemd/system/binfmt-support.service; enabled; vendor preset: enabled)
>>>>    Active: failed (Result: exit-code) since Sun 2020-05-10 21:54:27 CEST; 10h ago
>>>>      Docs: man:update-binfmts(8)
>>>>   Process: 353 ExecStart=/usr/sbin/update-binfmts --enable (code=exited, status=2)
>>>>  Main PID: 353 (code=exited, status=2)
>>>>
>>>> May 10 21:54:27 debian update-binfmts[353]: update-binfmts: unable to open /var/lib/binfmts: No such file or directory
>>>>
>>>> Any help is appreciated.
>>> This should help your problem:
>>>
>>> mkdir /etc/systemd/system/binfmt-support.service.d
>>>
>>> cat > /etc/systemd/system/binfmt-support.service.d/override.conf << EOF
>>> [Unit]
>>> RequiresMountsFor=/var
>>> EOF
>>
>> As another alternative, one can run "systemctl edit
>> binfmt-support.service", which will create the intervening folders and
>> files for you, and reload the daemon if the editor exits with success.
> 
> Thanks for suggestion! I have tried the advise and it actually worked
> (I will keep an eye on that because one reboot may not be representative).
> I wonder nevertheless what is the problem with this specific unit? It has
> dependency on local-fs.target which in turn should mount /var. So what
> exactly went wrong?


-- 
With best regards,
Dmitry