Re: [Exim] Override server callout verification is done on?

トップ ページ
このメッセージを削除
このメッセージに返信
著者: Stephen Gran
日付:  
To: exim-users
題目: Re: [Exim] Override server callout verification is done on?
On Wed, Jun 23, 2004 at 01:35:02AM -0400, Eli said:
> I have possibly a weird wishlist item - the ability to specify a server that
> a verification callout is sent to.
>
> What I mean by this is rather than have Exim do a DNS lookup, get the MX
> record, then connect and try to do a sender/recip verification callout, I'd
> like to just specify a single host/IP that would be used for the check.
>
> The reason for this is to have Exim as a front end to a backend mail server,
> but Exim won't have any knowledge of the userbase, so I want to do a callout
> to the backend system to verify if the user account exists or not. I know
> it's not the best of set ups, but maybe this may even have other uses better
> than this scenario?
>
> Just wondering if it's even worth requesting, or if there are too many
> issues that could arise from having such an ability?


We do this at one of the ISP's we service. We have multiple front-end
machines that recieve mail, and then deliver to the pop machine behind
them. We do this:

domainlist relay_to_domains = lsearch;CONFDIR/mailertable

Where that file is of the form:

example.com: mail.example.com
example.net: mail.example.com
example.org: mail.example.com
mail.example.com: mail.example.com
mail.example.net: mail.example.com
mail.example.org: mail.example.com

In the rcpt acl:

accept domains = +relay_to_domains
       endpass
       message = unrouteable address
       verify = recipient/callout=30s,defer_ok,no_cache,use_sender


first router:

relay_manualroute:
driver = manualroute
domains = +relay_to_domains
transport = remote_smtp
route_data = ${lookup{$domain}lsearch{CONFDIR/mailertable}}

and retry times:
example.com           *           F,6h,5m; G,16h,30m,1.5; F,6d,6h


The above does a pretty good job. If the backend pop server's load goes
too high, it will occasionally start rejecting connections, which is why
we use defer_ok and the 5 minute queue time. We also start exim with
-q5m to help this. The no_cache and use_sender are there so that the
backend machine can maintain per-user blacklists and so forth, and exim
doesn't get confused by mary@??? rejecting mail from
foo@???, but accepting everything else. We have seen that make exim
think mary@??? doesn't exist, so we went to the no_cache
solution - slightly more network traffic, but less overall headache.

HTH,
--
--------------------------------------------------------------------------
|  Stephen Gran                  | BOFH excuse #286:  Telecommunications   |
|  steve@???             | is downgrading.                         |
|  http://www.lobefin.net/~steve |                                         |

--------------------------------------------------------------------------