banshee/services-collection/services/postfix/ssh/queue

Postfix queue, by SSH

Returns the mail queue of a Postfix server: the messages queued, by queue (deferred, being delivered, held, arriving), and the age of the oldest deferred message, read by postqueue -j over SSH. Four watches are offered, each one set on the probe.

Versions

Postfix 3.1 and later (postqueue -j): Debian 9, Ubuntu 18.04, RHEL 8 and later. Generic: no folder of version. An older Postfix knows no -j: the collector stops misconfigured.

How it works

The collector runs postqueue -j, which writes one JSON object per message, one per line:

{"queue_name": "deferred", "queue_id": "4BPm4N2tVLz1x", "arrival_time": 1793062803, "message_size": 2048, "sender": "app@www.example", "recipients": [{"address": "user@mail.example", "delay_reason": "connect to mail.example[192.0.2.25]:25: Connection timed out"}]}

queued counts every message; deferred, active, hold count those of each queue; incoming counts those of incoming and maildrop, the mail arriving and not yet taken up.
oldest-deferred is the time since the arrival of the oldest deferred message, in hours, against the poller's clock; 0 when none is deferred. Postfix gives up on a message after maximal_queue_lifetime (5 days by default) and bounces it: a message deferred for hours is mail that does not leave.
A Postfix that is stopped (postqueue says "mail system is down") is critical: "Postfix is not running". Any other failure of postqueue stops the collector misconfigured, with what it said.
Nothing is written: postqueue -j only lists the queue. The senders and recipients are read, never returned.

Prerequisites

An SSH account for monitoring, without write rights, on the Postfix server.
postqueue in its path (/usr/sbin/postqueue), and the account allowed to list the queue: authorized_mailq_users is static:anyone by default; where it was narrowed, add the account to it (postconf -e 'authorized_mailq_users = monitoring', with the others it names).
No sudo.
The fingerprint of the server's host key, set in protocol:ssh:fingerprint (not needed when the Security settings trust the first SSH fingerprint presented).

Outputs

Name	Label	Unit	Per instance
`queued`	Messages queued	—	no
`deferred`	Messages deferred	—	no
`active`	Messages being delivered	—	no
`hold`	Messages held	—	no
`incoming`	Messages arriving	—	no
`oldest-deferred`	Oldest deferred message	h	no

Watches

Name	Title	Kind	Defaults	Checked by default
`queued`	Queue too long	above a threshold, on `queued`	warning above 500, critical above 2000	yes
`deferred`	Many messages deferred	above a threshold, on `deferred`	warning above 50, critical above 200, for 3 runs	yes
`oldest-deferred`	Deferred message too old	above a threshold, on `oldest-deferred`	warning above 4 h, critical above 24 h	yes
`hold`	Messages held	above a threshold, on `hold`	warning above 0	no

Limits

The whole queue is listed at each run: a queue of tens of thousands of messages makes an answer of several megabytes, which the poller may refuse as too large; the probe then stops, which says the queue is huge.
One Postfix instance per probe, the default one; the other instances of postmulti are not read.
Not checked against a Postfix: the format is the one the manual page postqueue(1) gives for -j.