[prev in list] [next in list] [prev in thread] [next in thread] 

List:       postfix-users
Subject:    Re: How does smtp_destination_concurrency_limit and smtp_destination_rate_delay relate?
From:       Rodrigo_Severo_-_Fábrica <rodrigo () fabricadeideias ! com>
Date:       2018-08-27 14:04:42
Message-ID: CAOdi6iaB476+=qhYA2BcnkfoMmDurEy2ZeGPwSQ5PF8PJ49NJA () mail ! gmail ! com
[Download RAW message or body]

On Mon, Aug 27, 2018 at 10:53 AM, Wietse Venema <wietse@porcupine.org> wrote:
> Rodrigo Severo - F?brica:
>> On Sun, Aug 26, 2018 at 2:15 PM, Wietse Venema <wietse@porcupine.org> wrote:
>> >
>> > Matus UHLAR - fantomas:
>> > > >Rodrigo Severo - F?brica:
>> > > >> I think that this info - the delay between deliveries is per
>> > > >> connection and not per, or only, per domain -  should be stated
>> > > >> clearly in the documentation.
>> > >
>> > > On 26.08.18 11:08, Wietse Venema wrote:
>> > > >Absolutely not. I promise that each delivery to that destination
>> > > >(recipient or domain) will be followed by the per-destination
>> > > >transport_destination_rate_delay.
>> > >
>> > > so, does transport_destination_concurrency_limit not apply when
>> > > transport_destination_rate_delay is >0 ?
>> >
>> > There can be no concurrency.
>> >
>> > With rate delay, there is an N second delay between the completion
>> > of delivery 1 to the destination, and the start of delivery 2 to
>> > that same destination.
>> >
>> > From this it follows that delivery 2 does not start before the rate
>> > delay expires, that delivery 2 does not start before delivery 1 has
>> > completed, and that there is no overlap in time between delivery
>> > 1, 2, 3, 4, 5, 6, 7, 8, and so on, to that same destination.
>> >
>> > You can replace 'email' with 'car', and 'destinion' with 'bridge'.
>> >
>> > Imagine counting cars that pass over a bridge. Imagine that car 1
>> > starts crossing the bridge, and that some time later it arrives at
>> > the other side. Then there is an N second delay before car 2 starts
>> > crossing that same bridge.
>> >
>> > From this it follows that car 2 does not start crossing the bridge
>> > before the delay has expired, that car 2 does not start crossing
>> > the bridge before car 1 has arrived at the other side, and that
>> > there is no overlap in time that cars 1, 2, 3, 4, 5, 6, 7, 8, and
>> > so on, are crossing that same bridge.
>>
>> To use your analogy, my doubt is: if a bridge is a destination,
>> increasing transport_destination_concurrency_limit won't create new
>> lanes on this same bridge? Increasing
>> transport_destination_concurrency_limit won't make more cars pass at
>> the same time on this same bridge?
>
> Let me repeat the text from above: there is no overlap in time that
> cars 1, 2, 3, 4, 5, 6, 7, 8, and so on, are crossing that same
> bridge.
>
> To state the obviuos, there can be no more than one car on the bridge.

Ok, so setting  transport_destination_rate_delay > 0 disables
transport_destination_concurrency_limit (or fixes it at 1). I believe
this info could, and should, be stated clearly on the documentation.

Thanks for your help, patience and support.


Regards,

Rodrigo Severo
[prev in list] [next in list] [prev in thread] [next in thread]