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
