thank you again! Am 27.04.2011 17:47, schrieb Victor Duchovni: > On Wed, Apr 27, 2011 at 05:32:28PM +0200, Reindl Harald wrote: > >> ok, if there is no easy option if will leave it in peace >> because my settings below are "friendly" enough i mean >> >> initial_destination_concurrency = 5 >> smtp_destination_concurrency_limit = 5 >> default_destination_recipient_limit = 15 > > Why so few recipients per message? This just increases the total > message load when a message has many recipients...
Because with 30 one of the biggest ISP did reject a message
with tempfail the whole day and after setting to 15 it the
message was accpeted
the common use of this machine is our own newsletter-system
with bounce-managment and this does never add more than
one rcpt for several reasons, as said the main target is not
getting messages fast out - two reasons:
* it is more important that they are accepted on the first try
* the WAN-network should be left in peace for webservers
>> default_destination_concurrency_limit = 5
>> default_destination_concurrency_failed_cohort_limit = 5
>> default_destination_rate_delay = 1
>
> The concurrency limit is ignored when a rate delay is specified.
> You should probably change just the "smtp_" settings, not the
> "default_" settings. Also replace "initial_destination_concurrency" with
> "smtp_initial_destination_concurrency"
is there anything in the settings below now what you would change
with keep in mind that outgoing mails should not make angry admins
on big hosters resulting in rate-controls / blacklisting
body_checks_size_limit = 65535
queue_run_delay = 240
maximal_queue_lifetime = 2d
bounce_queue_lifetime = 2d
minimal_backoff_time = 900
maximal_backoff_time = 5400
message_size_limit = 36700160
max_idle = 60
in_flow_delay = ${stress?2}${stress:0}s
smtp_destination_recipient_limit = 15
smtp_initial_destination_concurrency = 5
smtp_destination_concurrency_limit = 5
smtp_destination_concurrency_failed_cohort_limit = 5
smtp_destination_rate_delay = 1
smtp_helo_timeout = ${stress?45}${stress:180}s
smtp_mail_timeout = ${stress?45}${stress:180}s
smtp_connect_timeout = ${stress?15}${stress:45}s
smtpd_error_sleep_time = ${stress?1}${stress:2}s
smtpd_soft_error_limit = ${stress?2}${stress:5}
smtpd_hard_error_limit = ${stress?5}${stress:10}
smtpd_peername_lookup = no
we are really trying the get the lists of our customers clean with
daily parsing of the logfiles and removing non-existent rcpts the
first time they senn from ALL tables, never send any messages to rcpts
on the "robinson-list" (even if double-opt-in) and we will not deliver
messages with an attitude we do not want receiving from others while
also targeting a low network-load the keep ressources for time-critical
services
"smtpd_peername_lookup" is not needed here because only two
IPs from "mynetworks" and the host-IP are allowed to connect
>> transport_retry_time = 30
>
> I've never seen anyone changing this parameter. You should should not
> have failing transports, and should not need to change this parameter
ok, i will remove this, see above
signature.asc
Description: OpenPGP digital signature
