On Wed, Nov 29, 2017 at 02:53:20PM +0530, an...@isac.gov.in wrote:
> > Well, the main issue here is the consistent 10s delay to complete
> > a delivery after 0s connection latency. I would guess that's some
> > sort of timeout on the relay end resolving the IP address of the
> > connecting client. With 10s to complete each delivery, your
> > throughput is not surprisingly quite low.
>
> Thank you so much for the debugging. We have a PIX sitting between
> client and relay host and what I could see as part of default
> configuration that, smtp_pix_workaround_delay_time is 10s. Could PIXbe a
> factor in delivery latency?
The PIX does not introduce a transaction delay. Rather, Postfix
implements some PIX-specific work-arounds for that have been in
the queue long enough. This is not related to your throughput
issue, but as Wietse says, do configure your PIX to not mess
with SMTP, its SMTP features are just harmful.
> I don't see 10s delay during normal operations, but I do see "enabling
> PIX workarounds: disable_esmtp delay_dotcrlf for" lines in the logs
> for the relay in question.
>
> Do you think, PIX could be the reason?
No. I've already explained you have a downstream latency problem,
either DNS or misconfigured artificial delays. An overloaded
system would have a more variable delay. Exactly 10s for each
delivery must be a timer expiring, which could be a DNS timeout
or a deliberate pause.
--
Viktor.
