On Tue, Jan 18, 2005 at 10:56:22AM +0000, Stefano Catani wrote:
> here is the message:
> http://mail.units.it/6474
>
> it contains a lot of email addresses and stops our mailserver
>
> these are the times on a dual PIII 1GHz (SpamAssassin 3.0.2)
>
> time spamc < 6474
> real 9m59.995s
> user 0m0.000s
> sys 0m0.000s
similar result here:
real 10m0.067s
user 0m0.010s
sys 0m0.000s
single PIII 1GHz 750MB SA 3.0.0
spamd (according to top) does not eat significantly CPU.
I called strace on the spamd process:
...
select(0, NULL, NULL, NULL, {1, 20000}) = 0 (Timeout)
open("/etc/protocols", O_RDONLY) = -1 EMFILE (Too many open files)
open("/var/lib/misc/protocols.db", O_RDWR|O_LARGEFILE) = -1 EMFILE (Too many
open files)
...
this is reported endlessly
so there seems to be a file handle problem.
According to lsof:
lsof | grep ^spamd | awk '{print $1,$2}' | sort | uniq -c
NrOF PID
37 spamd 20696
126 spamd 20698
129 spamd 20699
130 spamd 20700
1055 spamd 20701
38 spamd 26284
This surely is insane.
Process 20701 which is the actually scanning child process
has openend 933 UDP sockets:
spamd 20701 root 1023u IPv4 555058 UDP
*:38796
and 85 handles on bayes_toks:
spamd 20701 root 136u REG 58,2 5226496 656011
/home/chris/.spamassassin/bayes_toks
I'd guess the UDP sockets are from DNS lookups f. sender verify.
HTH, Chris
--
Christian Recktenwald : :
citecs GmbH : [EMAIL PROTECTED]
Unternehmensberatung fuer : voice +49 711 601 2090 : Boeblinger Strasse 189
EDV und Telekommunikation : fax +49 711 601 2092 : D-70199 Stuttgart
