You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Andrzej Adam Filip <an...@onet.eu> on 2008/08/11 12:01:49 UTC

Re: High load average [sendmail+milter]

"Martin Schweizer" <sc...@gmail.com> wrote:
> On my mail server, I run FreeBSD 7.0 with sendmail/spamass-milter.
> Since some times I see some high usage of the processes
> "spamass-milter". Since then it seems that I have now I bad response
> time. But I have no idea from when... I solutions could be to you two
> servers: one with spamassins and the other with sendmail.
> Do you can give me any hints?
>
> Regards
>
> # top
> last pid: 59299;  load averages:  7.94,  8.33,  8.33
> 140 processes: 9 running, 131 sleeping
> CPU states: 18.4% user,  0.0% nice, 81.6% system,  0.0% interrupt,  0.0% idle
> Mem: 279M Active, 21M Inact, 171M Wired, 15M Cache, 111M Buf, 511M Free
> Swap: 2023M Total, 2008K Used, 2021M Free
>
>   PID USERNAME  THR PRI NICE   SIZE    RES STATE    TIME   WCPU COMMAND
> 94761 root        1 128    0 12964K  3060K RUN     33.8H 11.82% spamass-milter
>   585 root        1 128    0 12964K  3056K RUN     33.0H 11.82% spamass-milter
> 99155 root        1 128    0 12964K  3060K RUN     33.2H 11.72% spamass-milter
> 89596 root        1 128    0 11112K  2792K RUN     36.4H 11.67% spamass-milter
> 93129 root        1 128    0 11116K  2796K RUN     34.6H 11.57% spamass-milter
> 89464 root        1 128    0 11112K  2792K RUN     36.5H 11.52% spamass-milter
> 89010 root        1 128    0 11104K  2784K RUN     36.9H 11.47% spamass-milter
>   660 root        1 128    0 37032K 34636K RUN    191.8H 11.18% python2.4
> [snip]

0) How many message do you expect your MTA to handle in "peak hour"?
[ Most fine tuning advises for sendmail depend on it ]

1) Which set of DNSBL do you use?

Good set of DNSBL checks should "cut off" almost 70% of spam before
milter has a chance to consult spamassassin. Try to reject most of spam
at cost lower than incurred by spamassassin.

http://www.sendmail.org/m4/features.html#enhdnsbl
[ consider using zen.spamhaus.org and dnsbl.sorbs.net ]

It is possible to make sendmail fine tune set of DNSBL used based on
"country of origin" e.g. lax rules for "near by countries" and picky
mode for CN, KR, TW.
[requires generating sendmail.cf also from non sendmail.org m4 files]
[more hints upon request]

2) Have you considered using MIMEDefang.org milter?

2a) It can do more checks than merely consulting spamassassin
    e.g. anti-MS-virus checks and SPF records checks
2b) It does not run as super user

-- 
[pl>en: Andrew] Andrzej Adam Filip : anfi@onet.eu : anfi@xl.wp.pl
Maybe ain't ain't so correct, but I notice that lots of folks who ain't
using ain't ain't eatin' well.
  -- Will Rogers

Re: High load average [sendmail+milter]

Posted by Martin Schweizer <sc...@gmail.com>.
Hello Andrzej

2008/8/11 Andrzej Adam Filip <an...@onet.eu>:
> "Martin Schweizer" <sc...@gmail.com> wrote:
>> On my mail server, I run FreeBSD 7.0 with sendmail/spamass-milter.
>> Since some times I see some high usage of the processes
>> "spamass-milter". Since then it seems that I have now I bad response
>> time. But I have no idea from when... I solutions could be to you two
>> servers: one with spamassins and the other with sendmail.
>> Do you can give me any hints?
>>
>> Regards
>>
>> # top
>> last pid: 59299;  load averages:  7.94,  8.33,  8.33
>> 140 processes: 9 running, 131 sleeping
>> CPU states: 18.4% user,  0.0% nice, 81.6% system,  0.0% interrupt,  0.0% idle
>> Mem: 279M Active, 21M Inact, 171M Wired, 15M Cache, 111M Buf, 511M Free
>> Swap: 2023M Total, 2008K Used, 2021M Free
>>
>>   PID USERNAME  THR PRI NICE   SIZE    RES STATE    TIME   WCPU COMMAND
>> 94761 root        1 128    0 12964K  3060K RUN     33.8H 11.82% spamass-milter
>>   585 root        1 128    0 12964K  3056K RUN     33.0H 11.82% spamass-milter
>> 99155 root        1 128    0 12964K  3060K RUN     33.2H 11.72% spamass-milter
>> 89596 root        1 128    0 11112K  2792K RUN     36.4H 11.67% spamass-milter
>> 93129 root        1 128    0 11116K  2796K RUN     34.6H 11.57% spamass-milter
>> 89464 root        1 128    0 11112K  2792K RUN     36.5H 11.52% spamass-milter
>> 89010 root        1 128    0 11104K  2784K RUN     36.9H 11.47% spamass-milter
>>   660 root        1 128    0 37032K 34636K RUN    191.8H 11.18% python2.4
>> [snip]
>
> 0) How many message do you expect your MTA to handle in "peak hour"?
> [ Most fine tuning advises for sendmail depend on it ]
>
> 1) Which set of DNSBL do you use?
>
> Good set of DNSBL checks should "cut off" almost 70% of spam before
> milter has a chance to consult spamassassin. Try to reject most of spam
> at cost lower than incurred by spamassassin.
>
> http://www.sendmail.org/m4/features.html#enhdnsbl
> [ consider using zen.spamhaus.org and dnsbl.sorbs.net ]
>
> It is possible to make sendmail fine tune set of DNSBL used based on
> "country of origin" e.g. lax rules for "near by countries" and picky
> mode for CN, KR, TW.
> [requires generating sendmail.cf also from non sendmail.org m4 files]
> [more hints upon request]
>
> 2) Have you considered using MIMEDefang.org milter?
>
> 2a) It can do more checks than merely consulting spamassassin
>    e.g. anti-MS-virus checks and SPF records checks
> 2b) It does not run as super user

I will check it and will give you an update in next view days.

Regards,

-- 
Martin Schweizer
schweizer.martin@gmail.com
Tel.: +41 32 512 48 54 (VoIP)
Fax: +1 619 3300587