Knowledge base
max-rcpt-per-transaction
<domain>
How many addresses the node names to the receiver in one transfer. A message with a large list does not go out at once but in parts: the manual walks through this with its own example — five thousand recipients of one domain with a limit of three thousand are transferred twice, first three thousand, then the remaining two.
The message body is transferred anew each time. That is the whole cost of the setting. Cut the list into five pieces, and the receiver got the body five times: traffic grew fivefold, even though there is still one message. The manual says it directly: smaller values increase parallelism and reduce efficiency.
What you get in return. The pieces can travel over different connections at once — one message stops being one long transfer. On a large list this is the difference between “stuck on one connection” and “spread across ten.”
The directive has two names. max-rcpt-per-message in this same block is the very same limit; the manual says so directly, calling one an alias of the other. In its output the engine prints the second name, not this one.
For our sending the limit does not come into play at all: warm-up sends messages to one recipient each, and there is nothing to cut. The setting becomes meaningful on a node that receives messages with large lists — there it is set together with the number of messages per connection max-msg-per-connection, not separately.
Leave it alone while the list per message is small. With one or two recipients the limit plays no part. The conversation about it starts where a message has hundreds of addresses of one domain.
Count in bytes, not in messages. The estimate is simple: the body size multiplied by the number of pieces is what goes to the receiver. Halving the limit doubles the traffic, and that number has to be checked against volume ceilings, if they are set.
Lower it at the receiver’s request. Major receivers limit the number of addresses per transaction themselves and refuse the extra ones. Then the limit is set by their rule, and this is the only case when it is chosen for reasons other than speed.
# the list is cut into pieces of a hundred addresses — the body travels as many times <domain yahoo.com> max-rcpt-per-transaction 100 max-msg-per-connection 5 </domain>
▲
Common mistakes and problems
what people get wrong and how it ends3 mistakes▶
Lowering it for parallelism without counting traffic
§ 3.3.10.82
Setting both names at once, treating them as different settings
§ 3.3.10.83
Copying the number from the submission block
§ 3.3.10.82
- Syntax
- max-rcpt-per-transaction <integer>
- Values
- a positive integer
- Default
- 1000 § 3.3.10.82
- Applying
- reload, no restart needed
- Queue
- will not touch what is already queued
- Where to write it
- <domain>
- How many times
- once in the block
Related directives
3 relations
max-rcpt-per-message<domain>The same field under its second name — so says the manual. § 3.3.10.82
pmta show settings yahoo.com/mta-news-1
What else to read
3 articles
◇How the policy for a receiver is assembledSettings are assembled from the most general pattern to the most precise; the order of blocks in the file means nothing.◇Backoff mode — the queue's second personalityA queue that a receiver has started refusing switches to a parallel set of limits, retries and routes — ten directives that act only in this mode.◇Reputation monitoring inside the nodeThe engine itself watches the receiving servers’ replies and changes the queue mode on a pattern match — a reaction within the connection, not through telemetry.
- Cost of a mistake
- reversible: harms only while it is set
- Scenarios
- Troubleshoot deliverability · Tune the pace for a provider