Re: [gentoo-user] problem with mail server

[kashani]

Marcin Nis'kiewicz wrote:
> Hello
> I'm testing mail server with mysql backend. Generally it works quite
> well. But from time to time during testing, single mails can't be send
> because of smtp errors:
>
> in mail.log
> Feb  3 13:47:37 mail postfix/smtpd[28339]: NOQUEUE: reject: RCPT from
> unknown[ip]: 451 4.3.0 <user@××××××.org>:
> <mailto:piotrpol@××××××××××.pl> Temporary lookup failure;
> from=<user@××××××.org <mailto:piotrpol@××××××××××.pl>>
> to=<user@×××××××.org <mailto:piotrpol@××××××××××.pl>> proto=ESMTP
> helo=<domain.org <http://domain.org> >
> in mail.warn
> Feb  3 13:47:37 kurier4 postfix/trivial-rewrite[2438]: warning:
> transport_maps lookup failure
>
> when I check transport_map:
>
> postconf | grep transport_map
> address_verify_transport_maps = $transport_maps
> fallback_transport_maps =
> mailbox_transport_maps =
> mydestination = $myhostname, localhost.localdomain, $transport_maps
> proxy_read_maps = $local_recipient_maps $mydestination
> $virtual_alias_maps $virtual_alias_domains $virtual_mailbox_maps
> $virtual_mailbox_domains $relay_recipient_maps $relay_domains
> $canonical_maps $sender_canonical_maps $recipient_canonical_maps
> $relocated_maps $transport_maps $mynetworks
> transport_maps = mysql:/etc/mail/sql/mysql-transport.cf
> <http://mysql-transport.cf>
>
> my /etc/mail/sql/mysql-transport.cf <http://mysql-transport.cf> looks
> like that:
> user = postfix
> password = password
> dbname = maildb
> table = transport
> select_field = destination
> where_field = domain
> hosts = 127.0.0.1
>
> Generally I'm thinking that it could be mysql error - but there is
> nothing wrong in its error log...
> I set really big limit of concurrent connections
>
> max_user_connections = 1000
>
> So what can be wrong?
> Any ideas?
>
> Thank in progress for any help
> best regards
> nichu

I think you've got a couple of problems, but none of them individually 
jump at as the cause of your problems. However making these three 
changes together might help.

1. Turn your max_user_connections in Mysql down to something sane. 
Default is 100 which is fine unless you're also running a web app 
against the same Mysql instance.

2. Use proxy in your Mysql connections from Postfix.
Postfix can be configured to open a connection to Mysql and keep it 
open. Basically acts a connection pool and keep Postfix from opening 
hundreds of connections to Mysql on a very busy server. I recommend 
*always* using the proxy: statement anytime you're connecting to Mysql 
from Postfix. Your new transport_map statement will look like this.

transport_maps =  proxy:mysql:/etc/mail/sql/mysql-transport.cf

Generally you shouldn't be running into connection issues because you're 
hitting Mysql on localhost which means it'll default to a socket 
connection. It's possible that opening a new session is taking to too 
long occasionally and using proxy should alleviate that.

3. You're using Postfix 2.1 or earlier query syntax.
Hell it might even be Postfix 1.x syntax. This is the new syntax for 
Postfix 2.2 or better. This really isn't a problem, but the new syntax 
is far more powerful and suspect bugs that creep into the parser around 
old syntax aren't noticed or getting fixed.

user = postfix
password = password
hosts = localhost
dbname = maildb
query = SELECT destination FROM domain WHERE domain='%s'

I'm not sure what how-to you've been using, but I'd look at a few others 
to see some of the other options available. The one you're using seems 
to be pretty far out of date. While not wrong in any way it isn't taking 
full advantage of the last seven years of updates in Postfix.

kashani