On 2025/12/15 1:01 PM, Rich via Exim-users wrote:
> > > Are all of the batch dumped in by your CRM sitting undelivered?
> > > Or some proportion delivered and some stuck?
> > > Or nearly all delivered and only a few stuck?
> >
> > I could see them on mailq.
> Yes, but how many of them?
well, most of them: about 21.6k - after the 400 or so that did get
delivered.
While your choice of system load limit is up to you, 12 feels low.
The VM only has 2 cores; the drives are SSD, the fs is ext4. I could
consider throwing more resources at it, but it didn't seem to be working
that hard before - I'll do more monitoring now.
The bit I couldn't fathom was the load being high while no deliveries
occurring. Hence my thought that the routing phase might be chewing up
resources; or that I was simply starting too many runners [for 2 cpus].
You could experiment with the "-oY" cmdline option in combination with
"-qq"; it will disable some of those possibilities (at the cost of fast
> ramp-up of queue runners, so you'd want to keep that 10s period. For
> the normal case with -qq you don't need such a short period in most
> cases).
I'll leave -oY for later, and try the more basic changes first.
I do hope you're running a local caching resolver.
I am now, thanks ;-)
it might be worth experimenting with split-spool.
I'm already using that, thanks.
FYI: today's 22k mailing. 13k emails sent within 1 minute of being
queued and 98.5% emails sent within 20 mins of being queued (pretty
good); no abandoned runners due to load. This used:
Queue option: -q1m
deliver_queue_load_max = 20
queue_run_max = 35
I've configured now with
Queue option: -qq5m
deliver_queue_load_max = 20 (same)
queue_run_max = 5
I'll see how tomorrow's mailing goes.
Thanks,
--
Rich
--
## subscription configuration (requires account):
## https://lists.exim.org/mailman3/postorius/lists/exim-users.lists.exim.org/
## unsubscribe (doesn't require an account):
## [email protected]
## Exim details at http://www.exim.org/
## Please use the Wiki with this list - http://wiki.exim.org/
