Milter-greylist

Hello,

after succesfully using dnsrbl greylising, the last wave of heavy
spam showed logs where sendmail switched to default:
delivering spam.

Apr 17 16:35:01 mailhost sendmail[27825]: xyz: Milter (greylist): timeout before data read
Apr 17 16:35:01 mailhost sendmail[27825]: xyz: Milter (greylist): to error state

With no dnsrbl racl rule enabled milter-greylist-4.0a1 did not produce that
behaviour any more.

I understand that a DNS lookup may take too long (server not responding ...)
but then it would be nice if I could specify a timeout after
which milter-greylist would continue with the next racl line, and
it would be nice to know which sendmail configuration fits.

Another solution would be to throttle / disable dnsrbl lookups
when these take too long.

Any idea?

Michael Fromme

Michael Fromme <fromme@...> wrote:

> I understand that a DNS lookup may take too long (server not responding ...)
> but then it would be nice if I could specify a timeout after
> which milter-greylist would continue with the next racl line, and
> it would be nice to know which sendmail configuration fits.

We could have optionnal timeout and retry settings at the end of dnsrbl
definitions, for instance

dnsrbl "test" rbl.example.net 127.0.0.2/32 timeout 5s retry 3

After res_ninit has been called in dnsrbl.c, you'd just need to set
res.retrans and res.retry  according to the setting

Would you like to contribute that feature? It's quite straightforward.

-- 
Emmanuel Dreyfus
http://hcpnet.free.fr/pubz
manu@...