<domain>
The PIPELINING extension lets several commands be sent in a row without waiting for a reply to each. This noticeably speeds up the conversation, especially with many recipients in one message. The line cancels its use.
The manual gives the same reason as for ignore-chunking — debugging. When transfer behaves strangely, a one-command-at-a-time conversation reads unambiguously in the log: you can see exactly which command the receiver answered and how.
The price is speed. Without pipelining each command costs a full network round trip. For a message with a hundred recipients that is a hundred waits instead of a few.
Only for the duration of an investigation. And with a comment explaining why.
Together with ignore-chunking. Turning both off, you get the simplest possible conversation: command — reply, DATA, body, dot.
Remember the price. For this line it is measurable — speed, unlike the paired one, where only the transfer mechanism is lost.
# TEMPORARY: investigating transfer drops, 28.08 — remove afterwards <domain problem.example.com> ignore-chunking true ignore-pipelining true</domain>
what people get wrong and how it ends2 mistakes▶
our experience
our experience
0 relations
pmta show settings problem.example.com/vmta1 | grep -i ignore-pipelining grep -i 'PIPELINING' /var/log/pmta/pmta.log | tail -10
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.