[Pkg-clamav-devel] Bug#925475: clamav-freshclam.service is "inactive (dead)" after logrotate postrotate script execution

[Nils Fahldieck - Profihost AG]

Am 25.03.19 um 18:45 schrieb Adam D. Barratt:
> On 2019-03-25 15:39, Nils Fahldieck - Profihost AG wrote:
>> libc6 Version: 2.24-11+deb9u4
>> systemd Version: 232-25+deb9u9
>> Debian Version: Debian GNU/Linux 9 (stretch)
>>
>> The `postrotate` script in `/etc/logrotate.d/clamav-freshclam` executes
>> this code if systemd is used:
>>
>> $ cat /etc/logrotate.d/clamav-freshclam
>>      postrotate
>>      if [ -d /run/systemd/system ]; then
>>          systemctl -q is-active clamav-freshclam && systemctl kill
>> --signal=SIGHUP clamav-freshclam || true
>>
>> Whenever logrotate rotates freshclam's logfile, the service is inactive
>> (dead) afterwards.
>>
>> Reproducer:
>>
>> $ systemctl is-active clamav-freshclam
>> active
>> $ systemctl -q is-active clamav-freshclam && systemctl kill
>> --signal=SIGHUP clamav-freshclam || true
>> $ systemctl is-active clamav-freshclam
>> inactive
>> $ systemctl status clamav-freshclam.service
>> ● clamav-freshclam.service - ClamAV virus database updater
>>    Loaded: loaded (/lib/systemd/system/clamav-freshclam.service;
>> enabled; vendor preset: enabled)
>>   Drop-In: /etc/systemd/system/clamav-freshclam.service.d
>>            └─limits.conf
>>    Active: inactive (dead) since Mon 2019-03-25 16:28:25 CET; 2s ago
>>      Docs: man:freshclam(1)
> 
> fwiw, I can't reproduce this.
> 
> clamav-freshclam 0.100.2+dfsg-0+deb9u1
> libc6 2.24-11+deb9u4
> systemd 232-25+deb9u9
> 
> ● clamav-freshclam.service - ClamAV virus database updater
>    Loaded: loaded (/lib/systemd/system/clamav-freshclam.service;
> enabled; vendor preset: enabled)
>    Active: active (running) since Sun 2019-03-10 19:07:03 GMT; 2 weeks 0
> days ago
> ..
> 
> -rw-r----- 1 clamav adm   4754 Mar 11 06:25
> /var/log/clamav/freshclam.log.3.gz
> -rw-r----- 1 clamav adm   3544 Mar 17 06:25
> /var/log/clamav/freshclam.log.2.gz
> -rw-r----- 1 clamav adm 312142 Mar 24 06:25 /var/log/clamav/freshclam.log.1
> -rw-r----- 1 clamav adm  65671 Mar 25 17:25 /var/log/clamav/freshclam.log
> 
> also:
> 
> # systemctl is-active clamav-freshclam
> active
> # systemctl -q is-active clamav-freshclam && systemctl kill
> --signal=SIGHUP clamav-freshclam || true
> # systemctl is-active clamav-freshclam
> active
> 
> Aaaaaahhh... My guess would be that this is due to your local
> customisation:
> 
>  Drop-In: /etc/systemd/system/clamav-freshclam.service.d
>            └─limits.conf
> ...
>   Process: 15231 ExecStart=/usr/bin/cpulimit -f -l 5 --
> /usr/bin/freshclam -d --foreground=true (code=killed, signal=HUP)
> 
> which means that the HUP will be being sent to cpulimit, not to freshclam.
> 
> I'm not a clamav maintainer, but I'd be inclined at this point to say
> that this is not a problem with the package, and "you get to keep the
> pieces", I'm afraid. From a quick bit of research, I would suggest that
> you want to be using systemd's built-in
> https://www.freedesktop.org/software/systemd/man/systemd.resource-control.html#CPUQuota=
> , which would entirely avoid this issue.
> 
> Regards,
> 
> Adam

Hi Adam,

thank you very much for your answer and your time. You've guided me in
the right direction. In fact cpulimit receives the SIGHUP, which kills
the whole Cgroup of clamav-freshclam.

Regarding your suggestion to use CPUQuota: I will look more into
systemd's own options and use them the next time round.

So this is *not* a bug in clamav-freshclam package. I'm very sorry for
this false bug report.

Kind Regards,
Nils