QMGR(8)
π QMGR(8) β the Postfix queue manager, bestie edition π
System Manager's Manual Β· qmgr(8)
πΈ NAME πΈ
qmgr β the Postfix queue manager. She's the one running the whole delivery operation. main character energy fr π
β¨ SYNOPSIS β¨
qmgr [generic Postfix daemon options]
π¦ DESCRIPTION π¦
Okay so bestie, the qmgr(8) daemon just sits there and waits for incoming mail to show up, and then she arranges for it to get delivered by the Postfix delivery processes. She does not decide where stuff goes β the actual mail routing strategy is delegated to the trivial-rewrite(8) daemon. Not her job, she's just coordinating π
.
This program expects to be run from the master(8) process manager.
Mail that's addressed to the local double-bounce address? She logs it and discards it. And that's on purpose, bestie β that stops potential loops caused by undeliverable bounce notifications. Not me fixing the infinite bounce spiral ππ
π MAIL QUEUES π
The qmgr(8) daemon maintains these queues, and you're gonna learn every single one:
incomingβ Inbound mail from the network, or mail picked up by the localpickup(8)daemon from themaildropdirectory.activeβ Messages that the queue manager has opened for delivery. Only a limited number of messages is allowed to enter the active queue (leaky bucket strategy, for a fixed delivery rate). Gatekeeping, honestly βdeferredβ Mail that could not be delivered upon the first attempt. The queue manager implements exponential backoff by doubling the time between delivery attempts. She's not ghosting you, she's just waiting a lil longer each time πcorruptβ Unreadable or damaged queue files are moved here for inspection. It's giving quarantine π¦holdβ Messages that are kept "on hold" are kept here until someone sets them free. Prison but make it mail π
π DELIVERY STATUS REPORTS π
The qmgr(8) daemon keeps an eye on per-message delivery status reports in these directories. Each status report file has the same name as the corresponding message file β matchy matchy π«Ά:
bounceβ Per-recipient status information about why mail is bounced. These files are maintained by thebounce(8)daemon.deferβ Per-recipient status information about why mail is delayed. These files are maintained by thedefer(8)daemon.traceβ Per-recipient status information as requested with the Postfix"sendmail -v"or"sendmail -bv"command. These files are maintained by thetrace(8)daemon.
And the qmgr(8) daemon is the one responsible for asking the bounce(8), defer(8) or trace(8) daemons to send delivery reports. She's the group chat admin of delivery reports, periodt π
π STRATEGIES π
The queue manager implements a whole variety of strategies, for either opening queue files (input) or for message delivery (output). The range?? Iconic.
- leaky bucket β This strategy limits the number of messages in the active queue and prevents the queue manager from running out of memory under heavy load. Keeping it cute and contained πͺ£
- fairness β When the active queue has room, the queue manager takes one message from the incoming queue and one from the deferred queue. This prevents a large mail backlog from blocking the delivery of new mail. No queue gets to hog the aux π§
- slow start β This strategy eliminates "thundering herd" problems by slowly adjusting the number of parallel deliveries to the same destination. Not her sprinting at 100 right out the gate π
- round robin β The queue manager sorts delivery requests by destination. Round-robin selection prevents one destination from dominating deliveries to other destinations. Everybody gets a turn, no favourites π«
- exponential backoff β Mail that cannot be delivered upon the first attempt is deferred. The time interval between delivery attempts is doubled after each attempt. Doubling. Every time. She's patient but make it math π
- destination status cache β The queue manager avoids unnecessary delivery attempts by maintaining a short-term, in-memory list of unreachable destinations. Girl keeps receipts (in RAM) π§
- preemptive message scheduling β The queue manager attempts to minimize the average per-recipient delay while still preserving the correct per-message delays, using a sophisticated preemptive message scheduling. Sophisticated!! Say it with your chest π₯
β TRIGGERS β
On an idle system, the queue manager waits for the arrival of trigger events, or she waits for a timer to go off. A trigger is a one-byte message. Depending on the message received, the queue manager performs one of these actions (and the message is followed by the symbolic constant used internally by the software β keep both in mind for the exam bestie):
D(QMGR_REQ_SCAN_DEFERRED) β Start a deferred queue scan. If a deferred queue scan is already in progress, that scan will be restarted as soon as it finishes.I(QMGR_REQ_SCAN_INCOMING) β Start an incoming queue scan. If an incoming queue scan is already in progress, that scan will be restarted as soon as it finishes.A(QMGR_REQ_SCAN_ALL) β Ignore deferred queue file time stamps. The request affects the next deferred queue scan.F(QMGR_REQ_FLUSH_DEAD) β Purge all information about dead transports and destinations.W(TRIGGER_REQ_WAKEUP) β Wakeup call, This is used by the master server to instantiate servers that should not go away forever. The action is to start an incoming queue scan. (Yes that comma is in the original, no I will not be editing it π)
The qmgr(8) daemon reads an entire buffer worth of triggers. Multiple identical trigger requests are collapsed into one, and trigger requests are sorted so that A and F precede D and I. So, in order to force a deferred queue run, one would request A F D; in order to notify the queue manager of the arrival of new mail one would request I. Write that down. It's giving cheat code π
π STANDARDS π
π SECURITY π
The qmgr(8) daemon is not security sensitive. She reads single-character messages from untrusted local users, and thus may be susceptible to denial of service attacks. Not me getting spammed by randos π
The qmgr(8) daemon does not talk to the outside world, and she can be run at fixed low privilege in a chrooted environment. Cozy. Contained. Unbothered β¨
π DIAGNOSTICS π
Problems and transactions are logged to syslogd(8) or postlogd(8). Corrupted message files are saved to the corrupt queue for further inspection.
Depending on the setting of the notify_classes parameter, the postmaster is notified of bounces and of other trouble. Someone's getting pinged, bestie π
π BUGS π
A single queue manager process has to compete for disk access with multiple front-end processes such as cleanup(8). A sudden burst of inbound mail can negatively impact outbound delivery rates. One girl, so many disk-hungry friends, it's rough out here π«
π§ CONFIGURATION PARAMETERS π§
Changes to main.cf are not picked up automatically as qmgr(8) is a persistent process. Use the "postfix reload" command after a configuration change. Don't forget the reload, I'm begging π
The text below provides only a parameter summary. See postconf(5) for more details including examples.
In the text below, transport is the first field in a master.cf entry.
πΈ COMPATIBILITY CONTROLS πΈ
Available before Postfix version 2.5:
allow_min_user(no) β Allow a sender or recipient address to have`-`as the first character.
Available with Postfix version 2.7 and later:
default_filter_nexthop(empty) β When acontent_filterorFILTERrequest specifies no explicit next-hop destination, use$default_filter_nexthopinstead; when that value is empty, use the domain in the recipient address.
π ACTIVE QUEUE CONTROLS π
qmgr_clog_warn_time(300s) β The minimal delay between warnings that a specific destination is clogging up the Postfix active queue.qmgr_message_active_limit(20000) β The maximal number of messages in the active queue.qmgr_message_recipient_limit(20000) β The maximal number of recipients held in memory by the Postfix queue manager, and the maximal size of the short-term, in-memory "dead" destination status cache.qmgr_message_recipient_minimum(10) β The minimal number of in-memory recipients for any message.default_recipient_limit(20000) β The default per-transport upper limit on the number of in-memory recipients.transport_recipient_limit($default_recipient_limit) β A transport-specific override for thedefault_recipient_limitparameter value, where transport is themaster.cfname of the message delivery transport.default_extra_recipient_limit(1000) β The default value for the extra per-transport limit imposed on the number of in-memory recipients.transport_extra_recipient_limit($default_extra_recipient_limit) β A transport-specific override for thedefault_extra_recipient_limitparameter value, where transport is themaster.cfname of the message delivery transport.
Available in Postfix version 2.4 and later:
default_recipient_refill_limit(100) β The default per-transport limit on the number of recipients refilled at once.transport_recipient_refill_limit($default_recipient_refill_limit) β A transport-specific override for thedefault_recipient_refill_limitparameter value, where transport is themaster.cfname of the message delivery transport.default_recipient_refill_delay(5s) β The default per-transport maximum delay between refilling recipients.transport_recipient_refill_delay($default_recipient_refill_delay) β A transport-specific override for thedefault_recipient_refill_delayparameter value, where transport is themaster.cfname of the message delivery transport.
---β¨ DELIVERY CONCURRENCY CONTROLS β¨---
initial_destination_concurrency(5) β the starter-level per-destination concurrency for parallel deliveries all headed to the same destination. like, the opening number before the algorithm starts adjusting its little vibe πdefault_destination_concurrency_limit(20) β the default max number of parallel deliveries to the same destination. the ceiling, bestie.transport_destination_concurrency_limit($default_destination_concurrency_limit) β a transport-specific override for thedefault_destination_concurrency_limitvalue, where transport is themaster.cfname of the message delivery transport. so ifsmtpwants different rules than everyone else, this is the door πͺ- Available in Postfix version 2.5 and later:
transport_initial_destination_concurrency($initial_destination_concurrency) β a transport-specific override for theinitial_destination_concurrencyvalue, where transport is themaster.cfname of the message delivery transport.
default_destination_concurrency_failed_cohort_limit(1) β how many pseudo-cohorts have to eat a connection or handshake failure before that specific destination gets marked unavailable and further delivery gets suspended. it's giving "we tried, we're done" πtransport_destination_concurrency_failed_cohort_limit($default_destination_concurrency_failed_cohort_limit) β a transport-specific override for thedefault_destination_concurrency_failed_cohort_limitvalue, where transport is themaster.cfname of the message delivery transport.default_destination_concurrency_negative_feedback(1) β the per-destination amount of delivery concurrency negative feedback, applied after a delivery finishes with a connection or handshake failure. the algorithm takes the L and scales down π«Άtransport_destination_concurrency_negative_feedback($default_destination_concurrency_negative_feedback) β a transport-specific override for thedefault_destination_concurrency_negative_feedbackvalue, where transport is themaster.cfname of the message delivery transport.default_destination_concurrency_positive_feedback(1) β the per-destination amount of delivery concurrency positive feedback, applied after a delivery finishes without a connection or handshake failure. glow-up achieved, concurrency goes up β¨transport_destination_concurrency_positive_feedback($default_destination_concurrency_positive_feedback) β a transport-specific override for thedefault_destination_concurrency_positive_feedbackvalue, where transport is themaster.cfname of the message delivery transport.destination_concurrency_feedback_debug(no) β makes the queue manager's feedback algorithm verbose, for performance analysis purposes. turn this on and it will NOT stop yapping (affectionately) π£οΈ
πΈ RECIPIENT SCHEDULING CONTROLS πΈ
default_destination_recipient_limit(50) β the default max number of recipients per message delivery.transport_destination_recipient_limit($default_destination_recipient_limit) β a transport-specific override for thedefault_destination_recipient_limitvalue, where transport is themaster.cfname of the message delivery transport.
π MESSAGE SCHEDULING CONTROLS π
default_delivery_slot_cost(5) β how often the Postfix queue manager's scheduler is allowed to preempt delivery of one message with another. sharing is caring, but make it math π’transport_delivery_slot_cost($default_delivery_slot_cost) β a transport-specific override for thedefault_delivery_slot_costvalue, where transport is themaster.cfname of the message delivery transport.default_minimum_delivery_slots(3) β how many recipients a message must have in order to invoke the Postfix queue manager's scheduling algorithm at all. below that, the scheduler doesn't even look up π€transport_minimum_delivery_slots($default_minimum_delivery_slots) β a transport-specific override for thedefault_minimum_delivery_slotsvalue, where transport is themaster.cfname of the message delivery transport.default_delivery_slot_discount(50) β the default value for transport-specific_delivery_slot_discountsettings.transport_delivery_slot_discount($default_delivery_slot_discount) β a transport-specific override for thedefault_delivery_slot_discountvalue, where transport is themaster.cfname of the message delivery transport.default_delivery_slot_loan(3) β the default value for transport-specific_delivery_slot_loansettings.transport_delivery_slot_loan($default_delivery_slot_loan) β a transport-specific override for thedefault_delivery_slot_loanvalue, where transport is themaster.cfname of the message delivery transport.
π¦ OTHER RESOURCE AND RATE CONTROLS π¦
minimal_backoff_time(300s) β the minimal time between attempts to deliver a deferred message. heads up: prior to Postfix 2.4 the default value was1000s.maximal_backoff_time(4000s) β the maximal time between attempts to deliver a deferred message.maximal_queue_lifetime(5d) β consider a message undeliverable when delivery fails with a temporary error and the time in the queue has reached themaximal_queue_lifetimelimit. after that it's giving expired, bestie πͺ¦queue_run_delay(300s) β the time between deferred queue scans by the queue manager. also heads up: prior to Postfix 2.4 the default value was1000s.transport_retry_time(60s) β the time between attempts by the Postfix queue manager to contact a malfunctioning message delivery transport. giving it a lil breather before checking on it again π«Ά- Available in Postfix version 2.1 and later:
bounce_queue_lifetime(5d) β consider a bounce message undeliverable when delivery fails with a temporary error and the time in the queue has reached thebounce_queue_lifetimelimit.
- Available in Postfix version 2.5 and later:
default_destination_rate_delay(0s) β the default amount of delay inserted between individual message deliveries to the same destination and over the same message delivery transport. a lil pause for the plot βΈοΈtransport_destination_rate_delay($default_destination_rate_delay) β a transport-specific override for thedefault_destination_rate_delayvalue, where transport is themaster.cfname of the message delivery transport.
- Available in Postfix version 3.1 and later:
default_transport_rate_delay(0s) β the default amount of delay inserted between individual message deliveries over the same message delivery transport, regardless of destination. no destination exception, no cap, periodt βtransport_transport_rate_delay($default_transport_rate_delay) β a transport-specific override for thedefault_transport_rate_delayvalue, where the initial transport in the parameter name is themaster.cfname of the message delivery transport. yes the name says transport twice, that's not a typo, that's just how she is π
π SAFETY CONTROLS π
qmgr_daemon_timeout(1000s) β how much time a Postfix queue manager process may take to handle a request before a built-in watchdog timer terminates it. the watchdog does not play πqmgr_ipc_timeout(60s) β the time limit for the queue manager to send or receive information over an internal communication channel.- Available in Postfix version 3.1 and later:
address_verify_pending_request_limit(seepostconf -doutput) β a safety limit that stops address verification requests from overwhelming the Postfix queue. protect the queue at all costs, she's delicate π₯Ί
β MISCELLANEOUS CONTROLS β
config_directory(seepostconf -doutput) β the default location of the Postfixmain.cfandmaster.cfconfiguration files.defer_transports(empty) β the names of message delivery transports that should not deliver mail unless someone issues"sendmail -q"or equivalent. girl is on do-not-disturb π΄delay_logging_resolution_limit(2) β the maximal number of digits after the decimal point when logging delay values.helpful_warnings(yes) β log warnings about problematic configuration settings, and provide helpful suggestions. she's a supportive bestie fr fr π«Άprocess_id(read-only) β the process ID of a Postfix command or daemon process.process_name(read-only) β the process name of a Postfix command or daemon process.queue_directory(seepostconf -doutput) β the location of the Postfix top-level queue directory.syslog_facility(mail) β the syslog facility of Postfix logging.syslog_name(seepostconf -doutput) β a prefix that is prepended to the process name in syslog records, so that, for example,"smtpd"becomes"prefix/smtpd". glow-up via prefix β¨- Available in Postfix version 3.0 and later:
confirm_delay_cleared(no) β after sending a"your message is delayed"notification, inform the sender when the delay clears up. closure! we love closure π
- Available in Postfix 3.3 and later:
service_name(read-only) β themaster.cfservice name of a Postfix daemon process.
- Available in Postfix 3.5 and later:
info_log_address_format(external) β the email address form that will be used in non-debug logging (info,warning, etc.).
π FILES π
/var/spool/postfix/incoming, incoming queue/var/spool/postfix/active, active queue/var/spool/postfix/deferred, deferred queue/var/spool/postfix/bounce, non-delivery status/var/spool/postfix/defer, non-delivery status/var/spool/postfix/trace, delivery status
π SEE ALSO π
trivial-rewrite(8), address routingbounce(8), delivery status reportspostconf(5), configuration parametersmaster(5), generic daemon optionsmaster(8), process managerpostlogd(8), Postfix loggingsyslogd(8), system logging
π README FILES π
Use "postconf readme_directory" or "postconf html_directory" to locate this information.
SCHEDULER_README, scheduling algorithmQSHAPE_README, Postfix queue analysis
π LICENSE π
The Secure Mailer license must be distributed with this software. no skips, no exceptions, bestie π
π«Ά AUTHOR(S) π«Ά
Wietse Venema IBM T.J. Watson Research P.O. Box 704 Yorktown Heights, NY 10598, USA
Preemptive scheduler enhancements:
Patrik Rak Modra 6 155 00, Prague, Czech Republic
Wietse Venema Google, Inc. 111 8th Avenue New York, NY 10011, USA
and that's a wrap on the whole page π
the footer at the very bottom names this one as the QMGR(8) manual page β part 2 of 2, in the books, ate and left no crumbs πβ¨
π¦ where this page comes from
- alpine 3.24: postfix-doc 3.11.7-r0 Β· IPL-1.0 EPL-2.0
- arch rolling: postfix 3.11.7-1 Β· EPL-2.0 OR IPL-1.0 AND BSD-4-Clause-UC
- nixos 26.05: postfix 3.11.7
- nixpkgs unstable: postfix 3.11.7
- ubuntu stonking: postfix 3.11.7-1ubuntu2
- π girlified by deepseek-ai/DeepSeek-V4.1-Flash on : a language model's rewrite of the original, under its terms