Knowledge base

retry-interval verifiedimportantWritten by the Hub

Gaps between delivery attempts

in the configuration this is a lineretry-interval: 600,600,1800,1800,3600,3600,7200

Comma-separated list of waits between attempts: ten minutes twice, half an hour twice, an hour twice, then two hours.

Values

Values
comma-separated seconds, in order of attempt
Default
600,600,1800,1800,3600,3600,7200 · taken from the build
applying
restart required
where it can be set
in the config file · with a startup flag · as an environment variable

In detail

The list is read in order — first failure waits 600 seconds, second 600, third 1800, and so on — and the waiting itself is done by the broker's delay ladder, which composes each interval out of powers of two. Seven entries cover four hours and ten minutes in total, against a delivery deadline of twenty-four hours and an attempt limit of a hundred. What happens between the end of the list and the deadline is not stated anywhere in the package: either the last interval repeats, or attempts stop while the message waits out its deadline. That difference decides whether a temporarily refused message is retried nineteen more times or none, and it is the open question below. A per-message retries property overrides this list entirely, which is how a warm-up plan could give its own messages a gentler schedule than the node's default.

What happens if you get it wrong

Short intervals against a domain that is greylisting.Retries arrive before the greylist expires and are refused again, spending attempts without progress — the classic way to look like a spammer to a receiver that was merely cautious.inference · inference
Long intervals on time-sensitive mail.A password reset that waits two hours for its third attempt is useless when it arrives.inference · inference