https://bugs.exim.org/show_bug.cgi?id=1874
--- Comment #7 from David Carter <dpc22@???> ---
> You could look at max_parallel instead of serialize_hosts. But that isn't the
> issue...
Indeed, and I imagine that we would see the same problem if the rate limiting
was done by limiting concurrent connections at the Cyrus end.
> I assume you're hard-assigning users to mailstores, and splitting
> them out in a router? Could I see (the guts of) that?
localpartlist hermes_active = cdb;USERS/cyrus_active.cdb
Router:
# Deliver most Hermes addresses to Cyrus via LMTP. We don't need to do
# any address data handling at delivery time. The byname option causes
# Exim to skip the DNS lookup, which will fail anyway because the
# intramail names are not in the DNS, and which may cause delivery
# delays if the DNS does not work (e.g. because of a network outage).
# We work around interop problems in Cyrus by using an alternative
# transport if there are any zero bytes in the message.
hermes_lmtp:
driver = manualroute
local_part_suffix = +*
local_part_suffix_optional
no_verify
domains = hermes.cam.ac.uk
local_parts = +hermes_active
condition = SUBADDRESS_PARANOIA
route_data = $local_part_data byname
host_find_failed = defer
retry_use_local_part
transport = ${if ={0}{$body_zerocount} \
{hermes_lmtp} {hermes_lmtp_filter} }
Transports:
hermes_lmtp:
driver = smtp
protocol = lmtp
gethostbyname = true
rcpt_include_affixes = true
serialize_hosts = *
address_retry_include_sender = false
# This variant of the hermes_lmtp transport strips out any nul bytes in order
# to avoid triggering Cyrus's strict checking. We only use it when necessary
# for efficiency reasons.
hermes_lmtp_filter:
driver = smtp
protocol = lmtp
gethostbyname = true
rcpt_include_affixes = true
transport_filter = /usr/bin/tr -d \\000
serialize_hosts = *
address_retry_include_sender = false
--
You are receiving this mail because:
You are on the CC list for the bug.
