[pfx] Re: behavior of postscreen_dnsbl_min_ttl

2023-10-02 Thread Wietse Venema via Postfix-users
patpro--- via Postfix-users:
> hello,
> 
> September 29, 2023 4:30 PM, "Wietse Venema via Postfix-users" 
>  wrote:
> 
> > postscreen does not duplicate DNS caching. DNS lookup results are
> > already cached in a non-Postfix DNS resolver (see /etc/resolv.conf).
> 
> 
> OK, this is where I was wrong. I thought postscreen would cache
> the result of the DNS queries for at least postscreen_dnsbl_min_ttl.
> Most dnsbl have a crazy short TTL, I was hopping to get some cache
> here without having to install a tool capable of caching DNS query
> results with TTL override?

Postfix assumes that it will be using a caching resolver. Are you
using a non-caching resolver?

postscreen_dnsbl_min_ttl specifies how long POSTFIX will remeber
the result of a DNBL/WL test - the TEST RESULT, not the DNS response.

Wietse
___
Postfix-users mailing list -- postfix-users@postfix.org
To unsubscribe send an email to postfix-users-le...@postfix.org


[pfx] Re: behavior of postscreen_dnsbl_min_ttl

2023-10-02 Thread patpro--- via Postfix-users
hello,

September 29, 2023 4:30 PM, "Wietse Venema via Postfix-users" 
 wrote:

> postscreen does not duplicate DNS caching. DNS lookup results are
> already cached in a non-Postfix DNS resolver (see /etc/resolv.conf).


OK, this is where I was wrong. I thought postscreen would cache the result of 
the DNS queries for at least postscreen_dnsbl_min_ttl. Most dnsbl have a crazy 
short TTL, I was hopping to get some cache here without having to install a 
tool capable of caching DNS query results with TTL override…

May be I misunderstood this sentence in the documentation: «The minimum amount 
of time that postscreen(8) remembers that a client IP address passed a 
DNS-based reputation test […]»
In French «passer un test» means both «to take a test» and «to pass a test». As 
I understand it now, if a client fails the test (is denied), result is not 
cached. Is that correct?

 
> When postscreen receives multiple connections, then there can be
> multiple dnsblog queries.
> 
> Normally, postscreen will combine multiplw dnsblog queries for the
> same IP address into one query for that IP address, when connections
> from that IP address overlap in time during the PREGREET delay,
> but this client pregreets immediately (after 0.07s).
> 
> postscreen terminates the PREGREET delay as soon as the client
> pregreets and all dnsblog queries for that IP address have completed.
> That helps to get rid of spambots as quickly as possible.
> 
> But that also reduces the opportunities for connections to overlap,
> and thus, for multiplednsblog queries to be combined into one.


Thanks a lot for the detailed timing informations.

patpro
___
Postfix-users mailing list -- postfix-users@postfix.org
To unsubscribe send an email to postfix-users-le...@postfix.org