anvil

Tom Kinghorn

unread,

Jul 25, 2012, 9:02:49 AM7/25/12

to

Good afternoon

Apologies for the question but it is baffling me.

does Anvil record all connections?

I am receiving massive amounts of incoming mail (freakin status updates) from facebookmail.com and was hoping to check the connection rate via anvil in the logs.
However, there are no suck entries in the log.

I am trying to gets stats so that I can tweak:

smtpd_client_connection_rate_limit
smtpd_client_message_rate_limit
smtpd_client_recipient_rate_limit

many thanks and apologies for wasting your time on trivial questions.

Thanks
Tom

Wietse Venema

unread,

Jul 25, 2012, 10:07:30 AM7/25/12

to

Tom Kinghorn:

> Good afternoon
>
> Apologies for the question but it is baffling me.
> does Anvil record all connections?

As documented anvil counts the events within a time window. Once
the end of the time window is reached, the counters are reset to
zero and the next time window begins.

> I am receiving massive amounts of incoming mail (freakin status updates)
> from facebookmail.com and was hoping to check the connection rate via
> anvil in the logs.

As documented anvil is not a traffic management system. Instead,
it is a tool to slow down an out-of-control client until the
client is fixed or until it goes away,

If you get too much facebook email then anvil won't help,
because facebook is not an out-of-control client. All you're
achieving is a constant stream of retried deliveries.

Wietse

Stan Hoeppner

unread,

Jul 25, 2012, 5:00:02 PM7/25/12

to

On 7/25/2012 8:02 AM, Tom Kinghorn wrote:

> I am receiving massive amounts of incoming mail (freakin status updates)
> from facebookmail.com and was hoping to check the connection rate via
> anvil in the logs.

> However, there are no suck entries in the log.

> I am trying to gets stats so that I can tweak:
>
> smtpd_client_connection_rate_limit
> smtpd_client_message_rate_limit
> smtpd_client_recipient_rate_limit

I doubt you need to. I wouldn't bother with these. I'm not at all
familiar with Facebook's outbound systems, but if they act like most
other bulk mailers, each outbound host will open multiple concurrent
connections and may do connection caching on each. Simply set a low
value for:

smtpd_client_connection_count_limit

The default is maxproc/2 or 100/2=50. If you knock that down to
something like 2 or 4 it should decrease the Facebook load on your queue
substantially. A violator will generate this in your mail.warn log:

Jul 24 00:45:19 greer postfix/smtpd[4393]: warning: Connection
concurrency limit exceeded: 5 from unknown[115.153.143.13] for service smtp

My limit is 4.

Or you could use postfwd. But in practice I don't see what the
difference would be. Either way you're denying a client's connection
attempts. And since anvil is part of Postfix, and in machine code,
whereas postfwd is a perl policy daemon, it's must faster and more
efficient.

Others have very recently used smtpd_client_connection_count_limit
successfully to keep legit bulk senders from hammering their queues.
See the archives for the last few months.

--
Stan

Tom Kinghorn

unread,

Jul 26, 2012, 2:16:43 AM7/26/12

to

On 25/07/2012 23:00, Stan Hoeppner wrote:

On 7/25/2012 8:02 AM, Tom Kinghorn wrote:

I am trying to gets stats so that I can tweak:

smtpd_client_connection_rate_limit
smtpd_client_message_rate_limit
smtpd_client_recipient_rate_limit



Or you could use postfwd.  But in practice I don't see what the
difference would be.  Either way you're denying a client's connection
attempts.  And since anvil is part of Postfix, and in machine code,
whereas postfwd is a perl policy daemon, it's must faster and more
efficient.

Others have very recently used smtpd_client_connection_count_limit
successfully to keep legit bulk senders from hammering their queues.
See the archives for the last few months.

Many thanks to those who replied.

I understand now.

Regards
Tom