Configuration Reference

Description

Configuration files for Knot DNS use simplified YAML format. Simplified means that not all of the features are supported.

For the description of configuration items, we have to declare a meaning of the following symbols:

  • INT – Integer

  • STR – Textual string

  • HEXSTR – Hexadecimal string (with 0x prefix)

  • BOOL – Boolean value (on/off or true/false)

  • TIME – Number of seconds, an integer with possible time multiplier suffix (s ~ 1, m ~ 60, h ~ 3600 or d ~ 24 * 3600)

  • SIZE – Number of bytes, an integer with possible size multiplier suffix (B ~ 1, K ~ 1024, M ~ 1024^2 or G ~ 1024^3)

  • BASE64 – Base64 encoded string

  • ADDR – IPv4 or IPv6 address

  • DNAME – Domain name

  • ... – Multi-valued item, order of the values is preserved

  • [ ] – Optional value

  • | – Choice

The configuration consists of several fixed sections and optional module sections. There are 16 fixed sections (module, server, xdp, control, log, statistics, database, keystore, key, remote, remotes, acl, submission, policy, template, zone). Module sections are prefixed with the mod- prefix (e.g. mod-stats).

Most of the sections (e.g. zone) are sequences of settings blocks. Each settings block begins with a unique identifier, which can be used as a reference from other sections (such an identifier must be defined in advance).

A multi-valued item can be specified either as a YAML sequence:

address: [10.0.0.1, 10.0.0.2]

or as more single-valued items each on an extra line:

address: 10.0.0.1
address: 10.0.0.2

If an item value contains spaces or other special characters, it is necessary to enclose such a value within double quotes " ".

Comments

A comment begins with a # character and is ignored during processing. Also each configuration section or sequence block allows a permanent comment using the comment item which is stored in the server beside the configuration.

Including configuration

Another configuration file or files, matching a pattern, can be included at the top level in the current file.

include: STR

include

A path or a matching pattern specifying one or more files that are included at the place of the include option position in the configuration. If the path is not absolute, then it is considered to be relative to the current file. The pattern can be an arbitrary string meeting POSIX glob requirements, e.g. dir/*.conf. Matching files are processed in sorted order.

Default: not set

module section

Dynamic modules loading configuration.

Note

If configured with non-empty `--with-moduledir=path` parameter, all shared modules in this directory will be automatically loaded.

module:
  - id: STR
    file: STR

id

A module identifier in the form of the mod- prefix and module name suffix.

file

A path to a shared library file with the module implementation.

Warning

If the path is not absolute, the library is searched in the set of system directories. See man dlopen for more details.

Default: ${libdir}/knot/modules-${version}/module_name.so (or ${path}/module_name.so if configured with --with-moduledir=path)

server section

General options related to the server.

server:
    identity: [STR]
    version: [STR]
    nsid: [STR|HEXSTR]
    rundir: STR
    user: STR[:STR]
    pidfile: STR
    udp-workers: INT
    tcp-workers: INT
    background-workers: INT
    async-start: BOOL
    tcp-idle-timeout: TIME
    tcp-io-timeout: INT
    tcp-remote-io-timeout: INT
    tcp-max-clients: INT
    tcp-reuseport: BOOL
    tcp-fastopen: BOOL
    quic-max-clients: INT
    quic-outbuf-max-size: SIZE
    quic-idle-close-timeout: TIME
    remote-pool-limit: INT
    remote-pool-timeout: TIME
    remote-retry-delay: TIME
    socket-affinity: BOOL
    udp-max-payload: SIZE
    udp-max-payload-ipv4: SIZE
    udp-max-payload-ipv6: SIZE
    key-file: STR
    cert-file: STR
    edns-client-subnet: BOOL
    answer-rotation: BOOL
    automatic-acl: BOOL
    proxy-allowlist: ADDR[/INT] | ADDR-ADDR ...
    dbus-event: none | running | zone-updated | ksk-submission | dnssec-invalid ...
    dbus-init-delay: TIME
    listen: ADDR[@INT] ...

Caution

When you change configuration parameters dynamically or via configuration file reload, some parameters in the Server section require restarting the Knot server so that the changes take effect. See below for the details.

identity

An identity of the server returned in the response to the query for TXT record id.server. or hostname.bind. in the CHAOS class (RFC 4892). Set to an empty value to disable.

Default: FQDN hostname

version

A version of the server software returned in the response to the query for TXT record version.server. or version.bind. in the CHAOS class (RFC 4892). Set to an empty value to disable.

Default: server version

nsid

A DNS name server identifier (RFC 5001). Set to an empty value to disable.

Default: FQDN hostname at the moment of the daemon start

rundir

A path for storing run-time data (PID file, unix sockets, etc.).

Depending on the usage of this parameter, its change may require restart of the Knot server to take effect.

Default: ${localstatedir}/run/knot (configured with --with-rundir=path)

user

A system user with an optional system group (user:group) under which the server is run after starting and binding to interfaces. Linux capabilities are employed if supported.

Change of this parameter requires restart of the Knot server to take effect.

Default: root:root

pidfile

A PID file location.

Change of this parameter requires restart of the Knot server to take effect.

Default: rundir/knot.pid

udp-workers

A number of UDP workers (threads) used to process incoming queries over UDP.

Change of this parameter requires restart of the Knot server to take effect.

Default: equal to the number of online CPUs

tcp-workers

A number of TCP workers (threads) used to process incoming queries over TCP.

Change of this parameter requires restart of the Knot server to take effect.

Default: equal to the number of online CPUs, default value is at least 10

background-workers

A number of workers (threads) used to execute background operations (zone loading, zone updates, etc.).

Change of this parameter requires restart of the Knot server to take effect.

Default: equal to the number of online CPUs, default value is at most 10

async-start

If enabled, server doesn't wait for the zones to be loaded and starts responding immediately with SERVFAIL answers until the zone loads.

Default: off

tcp-idle-timeout

Maximum idle time (in seconds) between requests on an inbound TCP connection. It means if there is no activity on an inbound TCP connection during this limit, the connection is closed by the server.

Minimum: 1

Default: 10

tcp-io-timeout

Maximum time (in milliseconds) to receive or send one DNS message over an inbound TCP connection. It means this limit applies to normal DNS queries and replies, incoming DDNS, and outgoing zone transfers. The timeout is measured since some data is already available for processing. Set to 0 for infinity.

Default: 500 (milliseconds)

Caution

In order to reduce the risk of Slow Loris attacks, it's recommended setting this limit as low as possible on public servers.

tcp-remote-io-timeout

Maximum time (in milliseconds) to receive or send one DNS message over an outbound TCP connection which has already been established to a configured remote server. It means this limit applies to incoming zone transfers, sending NOTIFY, DDNS forwarding, and DS check or push. This timeout includes the time needed for a network round-trip and for a query processing by the remote. Set to 0 for infinity.

Default: 5000 (milliseconds)

tcp-reuseport

If enabled, each TCP worker listens on its own socket and the OS kernel socket load balancing is employed using SO_REUSEPORT (or SO_REUSEPORT_LB on FreeBSD). Due to the lack of one shared socket, the server can offer higher response rate processing over TCP. However, in the case of time-consuming requests (e.g. zone transfers of a TLD zone), enabled reuseport may result in delayed or not being responded client requests. So it is advisable to use this option on secondary servers.

Change of this parameter requires restart of the Knot server to take effect.

Default: off

tcp-fastopen

If enabled, use TCP Fast Open for outbound TCP communication (client side): incoming zone transfers, sending NOTIFY, and DDNS forwarding. This mode simplifies TCP handshake and can result in better networking performance. TCP Fast Open for inbound TCP communication (server side) isn't affected by this configuration as it's enabled automatically if supported by OS.

Note

The TCP Fast Open support must also be enabled on the OS level:

  • Linux/macOS: ensure kernel parameter net.ipv4.tcp_fastopen is 2 or 3 for server side, and 1 or 3 for client side.

  • FreeBSD: ensure kernel parameter net.inet.tcp.fastopen.server_enable is 1 for server side, and net.inet.tcp.fastopen.client_enable is 1 for client side.

Default: off

quic-max-clients

A maximum number of QUIC clients connected in parallel.

See also quic.

Change of this parameter requires restart of the Knot server to take effect.

Minimum: 128

Default: 10000 (ten thousand)

quic-outbuf-max-size

Maximum cumulative size of memory used for buffers of unACKed sent messages.

Note

Set low if little memory is available (together with quic-max-clients since QUIC connections are memory-heavy). Set to high value if outgoing zone transfers of big zone over QUIC are expected.

Change of this parameter requires restart of the Knot server to take effect.

Minimum: 1M (1 MiB)

Default: 100M (100 MiB)

quic-idle-close-timeout

Time in seconds, after which any idle QUIC connection is gracefully closed.

Change of this parameter requires restart of the Knot server to take effect.

Minimum: 1

Default: 4

remote-pool-limit

If nonzero, the server will keep up to this number of outgoing TCP connections open for later use. This is an optimization to avoid frequent opening of TCP connections to the same remote.

Change of this parameter requires restart of the Knot server to take effect.

Default: 0

remote-pool-timeout

The timeout in seconds after which the unused kept-open outgoing TCP connections to remote servers are closed.

Default: 5

remote-retry-delay

When a connection attempt times out to some remote address, this information will be kept for this specified time (in milliseconds) and other connections to the same address won't be attempted. This prevents repetitive waiting for timeout on an unreachable remote.

Default: 0

socket-affinity

If enabled and if SO_REUSEPORT is available on Linux, all configured network sockets are bound to UDP and TCP workers in order to increase the networking performance. This mode isn't recommended for setups where the number of network card queues is lower than the number of UDP or TCP workers.

Change of this parameter requires restart of the Knot server to take effect.

Default: off

tcp-max-clients

A maximum number of TCP clients connected in parallel, set this below the file descriptor limit to avoid resource exhaustion.

Note

It is advisable to adjust the maximum number of open files per process in your operating system configuration.

Default: one half of the file descriptor limit for the server process

udp-max-payload

Maximum EDNS0 UDP payload size default for both IPv4 and IPv6.

Default: 1232

udp-max-payload-ipv4

Maximum EDNS0 UDP payload size for IPv4.

Default: 1232

udp-max-payload-ipv6

Maximum EDNS0 UDP payload size for IPv6.

Default: 1232

key-file

Path to a server key PEM file which is used for DNS over QUIC communication.

Change of this parameter requires restart of the Knot server to take effect.

Default: one-time in-memory key

cert-file

Path to a server certificate PEM file which is used for DNS over QUIC communication.

Change of this parameter requires restart of the Knot server to take effect.

Default: one-time in-memory certificate

edns-client-subnet

Enable or disable EDNS Client Subnet support. If enabled, responses to queries containing the EDNS Client Subnet option always contain a valid EDNS Client Subnet option according to RFC 7871.

Default: off

answer-rotation

Enable or disable sorted-rrset rotation in the answer section of normal replies. The rotation shift is simply determined by a query ID.

Default: off

automatic-acl

If enabled, automatic ACL setting of configured remotes is considered when evaluating authorized operations.

Default: off

proxy-allowlist

An ordered list of IP addresses, network subnets, or network ranges which are allowed as a source address of proxied DNS traffic over UDP. The supported proxy protocol is haproxy PROXY v2.

Note

TCP is not supported.

Default: not set

dbus-event

Specification of server or zone states which emit a D-Bus signal on the system bus. The bus name is cz.nic.knotd, the object path is /cz/nic/knotd, and the interface name is cz.nic.knotd.events.

Possible values:

  • none – No signal is emitted.

  • running – The signal started is emitted when the server is fully operational and the signal stopped is emitted at the beginning of the server shutdown.

  • zone-updated – The signal zone_updated is emitted when a zone has been updated; the signal parameters are zone name and zone SOA serial.

  • ksk-submission – The signal zone_ksk_submission is emitted if there is a ready KSK present when the zone is signed; the signal parameters are zone name, KSK keytag, and KSK KASP id.

  • dnssec-invalid – The signal zone_dnssec_invalid is emitted when DNSSEC validation fails; the signal parameter is zone name.

Note

This function requires systemd version at least 221.

Change of this parameter requires restart of the Knot server to take effect.

Default: none

dbus-init-delay

Time in seconds which the server waits upon D-Bus initialization to ensure the D-Bus client is ready to receive signals.

Change of this parameter requires restart of the Knot server to take effect.

Minimum: 0

Default: 1

listen

One or more IP addresses where the server listens for incoming queries. Optional port specification (default is 53) can be appended to each address using @ separator. Use 0.0.0.0 for all configured IPv4 addresses or :: for all configured IPv6 addresses. Filesystem path can be specified for listening on local unix SOCK_STREAM socket. Non-local address binding is automatically enabled if supported by the operating system.

Change of this parameter requires restart of the Knot server to take effect.

Default: not set

xdp section

Various options related to XDP listening, especially TCP.

xdp:
    listen: STR[@INT] | ADDR[@INT] ...
    udp: BOOL
    tcp: BOOL
    quic: BOOL
    quic-port: INT
    quic-log: BOOL
    tcp-max-clients: INT
    tcp-inbuf-max-size: SIZE
    tcp-outbuf-max-size: SIZE
    tcp-idle-close-timeout: TIME
    tcp-idle-reset-timeout: TIME
    tcp-resend-timeout: TIME
    route-check: BOOL

Caution

When you change configuration parameters dynamically or via configuration file reload, some parameters in the XDP section require restarting the Knot server so that the changes take effect.

listen

One or more network device names (e.g. ens786f0) on which the Mode XDP is enabled. Alternatively, an IP address can be used instead of a device name, but the server will still listen on all addresses belonging to the same interface! Optional port specification (default is 53) can be appended to each device name or address using @ separator.

Change of this parameter requires restart of the Knot server to take effect.

Caution

If XDP workers only process regular DNS traffic over UDP, it is strongly recommended to also listen on the addresses which are intended to offer the DNS service, at least to fulfil the DNS requirement for working TCP.

Default: not set

udp

If enabled, DNS over UDP is processed with XDP workers.

Change of this parameter requires restart of the Knot server to take effect.

Default: on

tcp

If enabled, DNS over TCP traffic is processed with XDP workers.

The TCP stack limitations:

  • Congestion control is not implemented.

  • Lost packets that do not contain TCP payload may not be resend.

  • Not optimized for transfers of non-trivial zones.

Change of this parameter requires restart of the Knot server to take effect.

Default: off

quic

If enabled, DNS over QUIC is processed with XDP workers.

Change of this parameter requires restart of the Knot server to take effect.

Default: off

quic-port

DNS over QUIC will listen on the interfaces configured by listen, but on different port, configured by this option.

Change of this parameter requires restart of the Knot server to take effect.

Default: 853

quic-log

Triggers extensive logging of all QUIC protocol internals for every connection.

Change of this parameter requires restart of the Knot server to take effect.

Default: off

tcp-max-clients

A maximum number of TCP clients connected in parallel.

Minimum: 1024

Default: 1000000 (one million)

tcp-inbuf-max-size

Maximum cumulative size of memory used for buffers of incompletely received messages.

Minimum: 1M (1 MiB)

Default: 100M (100 MiB)

tcp-outbuf-max-size

Maximum cumulative size of memory used for buffers of unACKed sent messages.

Minimum: 1M (1 MiB)

Default: 100M (100 MiB)

tcp-idle-close-timeout

Time in seconds, after which any idle connection is gracefully closed.

Minimum: 1

Default: 10

tcp-idle-reset-timeout

Time in seconds, after which any idle connection is forcibly closed.

Minimum: 1

Default: 20

tcp-resend-timeout

Resend outgoing data packets (with DNS response payload) if not ACKed before this timeout.

Minimum: 1

Default: 5

route-check

If enabled, routing information from the operating system is considered when processing every incoming DNS packet received over the XDP interface:

  • If the outgoing interface of the corresponding DNS response differs from the incoming one, the packet is processed normally by UDP/TCP workers (XDP isn't used).

  • If the destination address is blackholed, unreachable, or prohibited, the DNS packet is dropped without any response.

  • The destination MAC address and possible VLAN tag for the response are taken from the routing system.

If disabled, symmetrical routing is applied. It means that the query source MAC address is used as a response destination MAC address. Possible VLAN tag is preserved.

Change of this parameter requires restart of the Knot server to take effect.

Note

This mode requires forwarding enabled on the loopback interface (sysctl -w net.ipv4.conf.lo.forwarding=1 and sysctl -w net.ipv6.conf.lo.forwarding=1). If forwarding is disabled, all incoming DNS packets are dropped!

Only VLAN 802.1Q is supported.

Default: off

control section

Configuration of the server control interface.

control:
    listen: STR
    timeout: TIME

listen

A UNIX socket path where the server listens for control commands.

Default: rundir/knot.sock

timeout

Maximum time (in seconds) the control socket operations can take. Set to 0 for infinity.

Default: 5

log section

Server can be configured to log to the standard output, standard error output, syslog (or systemd journal if systemd is enabled) or into an arbitrary file.

There are 6 logging severity levels:

  • critical – Non-recoverable error resulting in server shutdown.

  • error – Recoverable error, action should be taken.

  • warning – Warning that might require user action.

  • notice – Server notice or hint.

  • info – Informational message.

  • debug – Debug or detailed message.

In the case of a missing log section, warning or more serious messages will be logged to both standard error output and syslog. The info and notice messages will be logged to standard output.

log:
  - target: stdout | stderr | syslog | STR
    server: critical | error | warning | notice | info | debug
    control: critical | error | warning | notice | info | debug
    zone: critical | error | warning | notice | info | debug
    any: critical | error | warning | notice | info | debug

target

A logging output.

Possible values:

  • stdout – Standard output.

  • stderr – Standard error output.

  • syslog – Syslog or systemd journal.

  • file_name – A specific file.

With syslog target, syslog service is used. However, if Knot DNS has been compiled with systemd support and operating system has been booted with systemd, systemd journal is used for logging instead of syslog.

server

Minimum severity level for messages related to general operation of the server to be logged.

Default: not set

control

Minimum severity level for messages related to server control to be logged.

Default: not set

zone

Minimum severity level for messages related to zones to be logged.

Default: not set

any

Minimum severity level for all message types to be logged.

Default: not set

stats section

Periodic server statistics dumping.

statistics:
    timer: TIME
    file: STR
    append: BOOL

timer

A period after which all available statistics metrics will by written to the file.

Default: not set

file

A file path of statistics output in the YAML format.

Default: rundir/stats.yaml

append

If enabled, the output will be appended to the file instead of file replacement.

Default: off

database section

Configuration of databases for zone contents, DNSSEC metadata, or event timers.

database:
    storage: STR
    journal-db: STR
    journal-db-mode: robust | asynchronous
    journal-db-max-size: SIZE
    kasp-db: STR
    kasp-db-max-size: SIZE
    timer-db: STR
    timer-db-max-size: SIZE
    catalog-db: str
    catalog-db-max-size: SIZE

storage

A data directory for storing journal, KASP, and timer databases.

Default: ${localstatedir}/lib/knot (configured with --with-storage=path)

journal-db

An explicit specification of the persistent journal database directory. Non-absolute path (i.e. not starting with /) is relative to storage.

Default: storage/journal

journal-db-mode

Specifies journal LMDB backend configuration, which influences performance and durability.

Possible values:

  • robust – The journal database disk synchronization ensures database durability but is generally slower.

  • asynchronous – The journal database disk synchronization is optimized for better performance at the expense of lower database durability in the case of a crash. This mode is recommended on secondary servers with many zones.

Default: robust

journal-db-max-size

The hard limit for the journal database maximum size. There is no cleanup logic in journal to recover from reaching this limit. Journal simply starts refusing changes across all zones. Decreasing this value has no effect if it is lower than the actual database file size.

It is recommended to limit journal-max-usage per-zone instead of journal-db-max-size in most cases. Please keep this value larger than the sum of all zones' journal usage limits. See more details regarding journal behaviour.

Note

This value also influences server's usage of virtual memory.

Default: 20G (20 GiB), or 512M (512 MiB) for 32-bit

kasp-db

An explicit specification of the KASP database directory. Non-absolute path (i.e. not starting with /) is relative to storage.

Default: storage/keys

kasp-db-max-size

The hard limit for the KASP database maximum size.

Note

This value also influences server's usage of virtual memory.

Default: 500M (500 MiB)

timer-db

An explicit specification of the persistent timer database directory. Non-absolute path (i.e. not starting with /) is relative to storage.

Default: storage/timers

timer-db-max-size

The hard limit for the timer database maximum size.

Note

This value also influences server's usage of virtual memory.

Default: 100M (100 MiB)

catalog-db

An explicit specification of the zone catalog database directory. Only useful if Catalog zones are enabled. Non-absolute path (i.e. not starting with /) is relative to storage.

Default: storage/catalog

catalog-db-max-size

The hard limit for the catalog database maximum size.

Note

This value also influences server's usage of virtual memory.

Default: 20G (20 GiB), or 512M (512 MiB) for 32-bit

keystore section

DNSSEC keystore configuration.

keystore:
  - id: STR
    backend: pem | pkcs11
    config: STR
    key-label: BOOL

id

A keystore identifier.

backend

A key storage backend type.

Possible values:

  • pem – PEM files.

  • pkcs11 – PKCS #11 storage.

Default: pem

config

A backend specific configuration. A directory with PEM files (the path can be specified as a relative path to kasp-db) or a configuration string for PKCS #11 storage (<pkcs11-url> <module-path>).

Note

Example configuration string for PKCS #11:

"pkcs11:token=knot;pin-value=1234 /usr/lib64/pkcs11/libsofthsm2.so"

Default: kasp-db/keys

key-label

If enabled in combination with the PKCS #11 backend, generated keys are labeled in the form <zone_name> KSK|ZSK.

Default: off

key section

Shared TSIG keys used to authenticate communication with the server.

key:
  - id: DNAME
    algorithm: hmac-md5 | hmac-sha1 | hmac-sha224 | hmac-sha256 | hmac-sha384 | hmac-sha512
    secret: BASE64

id

A key name identifier.

Note

This value MUST be exactly the same as the name of the TSIG key on the opposite primary/secondary server(s).

algorithm

A TSIG key algorithm. See TSIG Algorithm Numbers.

Possible values:

  • hmac-md5

  • hmac-sha1

  • hmac-sha224

  • hmac-sha256

  • hmac-sha384

  • hmac-sha512

Default: not set

secret

Shared key secret.

Default: not set

remote section

Definitions of remote servers for outgoing connections (source of a zone transfer, target for a notification, etc.).

remote:
  - id: STR
    address: ADDR[@INT] ...
    via: ADDR[@INT] ...
    key: key_id
    block-notify-after-transfer: BOOL
    no-edns: BOOL
    automatic-acl: BOOL

id

A remote identifier.

address

An ordered list of destination IP addresses which are used for communication with the remote server. The addresses are tried in sequence until the remote is reached. Optional destination port (default is 53) can be appended to the address using @ separator.

Default: not set

Note

If the remote is contacted and it refuses to perform requested action, no more addresses will be tried for this remote.

via

An ordered list of source IP addresses. The first address with the same family as the destination address is used as a source address for communication with the remote. This option can help if the server listens on more addresses. Optional source port (default is random) can be appended to the address using @ separator.

Default: not set

key

A reference to the TSIG key which is used to authenticate the communication with the remote server.

Default: not set

block-notify-after-transfer

When incoming AXFR/IXFR from this remote (as a primary server), suppress sending NOTIFY messages to all configured secondary servers.

Default: off

no-edns

If enabled, no OPT record (EDNS) is inserted to outgoing requests to this remote server. This mode is necessary for communication with some broken implementations (e.g. Windows Server 2016).

Note

This option effectively disables zone expire timer updates via EDNS EXPIRE option specified in RFC 7314.

Default: off

automatic-acl

If enabled, some authorized operations for the remote are automatically allowed based on the context:

  • Incoming NOTIFY is allowed from the remote if it's configured as a primary server for the zone.

  • Outgoing zone transfer is allowed to the remote if it's configured as a NOTIFY target for the zone.

Automatic ACL rules are evaluated before explicit zone ACL configuration.

Note

This functionality requires global activation via automatic-acl in the server section.

Default: on

remotes section

Definitions of groups of remote servers. Remote grouping can simplify the configuration.

remotes:
  - id: STR
    remote: remote_id ...

id

A remote group identifier.

remote

An ordered list of references to remote server definitions.

Default: not set

acl section

Access control list rule definitions. An ACL rule is a description of one or more authorized operations (zone transfer request, zone change notification, and dynamic DNS update) which are allowed to be processed or denied.

acl:
  - id: STR
    address: ADDR[/INT] | ADDR-ADDR ...
    key: key_id ...
    remote: remote_id | remotes_id ...
    action: query | notify | transfer | update ...
    deny: BOOL
    update-type: STR ...
    update-owner: key | zone | name
    update-owner-match: sub-or-equal | equal | sub
    update-owner-name: STR ...

id

An ACL rule identifier.

address

An ordered list of IP addresses, network subnets, or network ranges. The query's source address must match one of them. If this item is not set, address match is not required.

Default: not set

key

An ordered list of references to TSIG keys. The query must match one of them. If this item is not set, transaction authentication is not used.

Default: not set

remote

An ordered list of references remote and remotes. The query must match one of the remotes. Specifically, one of the remote's addresses and remote's TSIG key if configured must match.

Note

This option cannot be specified along with the address or key option at one ACL item.

Default: not set

action

An ordered list of allowed (or denied) actions.

Possible values:

  • query – Allow regular DNS query. As normal queries are always allowed, this action is only useful in combination with TSIG key.

  • notify – Allow incoming notify (NOTIFY).

  • transfer – Allow zone transfer (AXFR, IXFR).

  • update – Allow zone updates (DDNS).

Default: query

deny

If enabled, instead of allowing, deny the specified action, address, key, or combination if these items. If no action is specified, deny all actions.

Default: off

update-type

A list of allowed types of Resource Records in a zone update. Every record in an update must match one of the specified types.

Default: not set

update-owner

This option restricts possible owners of Resource Records in a zone update by comparing them to either the TSIG key identity, the current zone name, or to a list of domain names given by the update-owner-name option. The comparison method is given by the update-owner-match option.

Possible values:

  • key — The owner of each updated RR must match the identity of the TSIG key if used.

  • name — The owner of each updated RR must match at least one name in the update-owner-name list.

  • zone — The owner of each updated RR must match the current zone name.

Default: not set

update-owner-match

This option defines how the owners of Resource Records in an update are matched to the domain name(s) set by the update-owner option.

Possible values:

  • sub-or-equal — The owner of each RR in an update must either be equal to or be a subdomain of at least one domain name set by update-owner.

  • equal — The owner of each updated RR must be equal to at least one domain name set by update-owner.

  • sub — The owner of each updated RR must be a subdomain of, but MUST NOT be equal to at least one domain name set by update-owner.

Default: sub-or-equal

update-owner-name

A list of allowed owners of RRs in a zone update used with update-owner set to name. Every listed owner name which is not FQDN (i.e. it doesn't end in a dot) is considered as if it was appended with the target zone name. Such a relative owner name specification allows better ACL rule reusability across multiple zones.

Default: not set

submission section

Parameters of KSK submission checks.

submission:
  - id: STR
    parent: remote_id | remotes_id ...
    check-interval: TIME
    timeout: TIME
    parent-delay: TIME

id

A submission identifier.

parent

A list of references remote and remotes to parent's DNS servers to be checked for presence of corresponding DS records in the case of KSK submission. All of them must have a corresponding DS for the rollover to continue. If none is specified, the rollover must be pushed forward manually.

Default: not set

Tip

A DNSSEC-validating resolver can be set as a parent.

check-interval

Interval for periodic checks of DS presence on parent's DNS servers, in the case of the KSK submission.

Default: 1h (1 hour)

timeout

After this time period (in seconds) the KSK submission is automatically considered successful, even if all the checks were negative or no parents are configured. Set to 0 for infinity.

Default: 0

parent-delay

After successful parent DS check, wait for this period before continuing the next key roll-over step. This delay shall cover the propagation delay of update in the parent zone.

Default: 0

policy section

DNSSEC policy configuration.

policy:
  - id: STR
    keystore: keystore_id
    manual: BOOL
    single-type-signing: BOOL
    algorithm: rsasha1 | rsasha1-nsec3-sha1 | rsasha256 | rsasha512 | ecdsap256sha256 | ecdsap384sha384 | ed25519 | ed448
    ksk-size: SIZE
    zsk-size: SIZE
    ksk-shared: BOOL
    dnskey-ttl: TIME
    zone-max-ttl: TIME
    ksk-lifetime: TIME
    zsk-lifetime: TIME
    delete-delay: TIME
    propagation-delay: TIME
    rrsig-lifetime: TIME
    rrsig-refresh: TIME
    rrsig-pre-refresh: TIME
    reproducible-signing: BOOL
    nsec3: BOOL
    nsec3-iterations: INT
    nsec3-opt-out: BOOL
    nsec3-salt-length: INT
    nsec3-salt-lifetime: TIME
    signing-threads: INT
    ksk-submission: submission_id
    ds-push: remote_id | remotes_id ...
    cds-cdnskey-publish: none | delete-dnssec | rollover | always | double-ds
    cds-digest-type: sha256 | sha384
    dnskey-management: full | incremental
    offline-ksk: BOOL
    unsafe-operation: none | no-check-keyset | no-update-dnskey | no-update-nsec | no-update-expired ...

id

A policy identifier.

keystore

A reference to a keystore holding private key material for zones.

Default: an imaginary keystore with all default values

Note

A configured keystore called "default" won't be used unless explicitly referenced.

manual

If enabled, automatic key management is not used.

Default: off

single-type-signing

If enabled, Single-Type Signing Scheme is used in the automatic key management mode.

Default: off (module onlinesign has default on)

algorithm

An algorithm of signing keys and issued signatures. See DNSSEC Algorithm Numbers.

Possible values:

  • rsasha1

  • rsasha1-nsec3-sha1

  • rsasha256

  • rsasha512

  • ecdsap256sha256

  • ecdsap384sha384

  • ed25519

  • ed448

Note

Ed25519 algorithm is only available if compiled with GnuTLS 3.6.0+.

Ed448 algorithm is only available if compiled with GnuTLS 3.6.12+ and Nettle 3.6+.

Default: ecdsap256sha256

ksk-size

A length of newly generated KSK or CSK keys.

Default: 2048 (rsa*), 256 (ecdsap256), 384 (ecdsap384), 256 (ed25519), 456 (ed448)

zsk-size

A length of newly generated ZSK keys.

Default: see default for ksk-size

ksk-shared

If enabled, all zones with this policy assigned will share one or more KSKs. More KSKs can be shared during a KSK rollover.

Warning

As the shared KSK set is bound to the policy id, renaming the policy breaks this connection and new shared KSK set is initiated when a new KSK is needed.

Default: off

dnskey-ttl

A TTL value for DNSKEY records added into zone apex.

Note

Has influence over ZSK key lifetime.

Warning

Ensure all DNSKEYs with updated TTL are propagated before any subsequent DNSKEY rollover starts.

Default: zone SOA TTL

zone-max-ttl

Declare (override) maximal TTL value among all the records in zone.

Note

It's generally recommended to override the maximal TTL computation by setting this explicitly whenever possible. It's required for DNSSEC Offline KSK and really reasonable when records are generated dynamically (e.g. by a module).

Default: computed after zone is loaded

ksk-lifetime

A period between KSK activation and the next rollover initiation.

Note

KSK key lifetime is also influenced by propagation-delay, dnskey-ttl, and KSK submission delay.

Zero (aka infinity) value causes no KSK rollover as a result.

This applies for CSK lifetime if single-type-signing is enabled.

Default: 0

zsk-lifetime

A period between ZSK activation and the next rollover initiation.

Note

More exactly, this period is measured since a ZSK is activated, and after this, a new ZSK is generated to replace it within following roll-over.

ZSK key lifetime is also influenced by propagation-delay and dnskey-ttl

Zero (aka infinity) value causes no ZSK rollover as a result.

Default: 30d (30 days)

delete-delay

Once a key (KSK or ZSK) is rolled-over and removed from the zone, keep it in the KASP database for at least this period before deleting it completely. This might be useful in some troubleshooting cases when resurrection is needed.

Default: 0

propagation-delay

An extra delay added for each key rollover step. This value should be high enough to cover propagation of data from the primary server to all secondary servers, as well as the duration of signing routine itself and possible outages in signing and propagation infrastructure. In other words, this delay should ensure that within this period of time after planned change of the key set, all public-facing secondaries will already serve new DNSKEY RRSet for sure.

Note

Has influence over ZSK key lifetime.

Default: 1h (1 hour)

rrsig-lifetime

A validity period of newly issued signatures.

Note

The RRSIG's signature inception time is set to 90 minutes in the past. This time period is not counted to the signature lifetime.

Default: 14d (14 days)

rrsig-refresh

A period how long at least before a signature expiration the signature will be refreshed, in order to prevent expired RRSIGs on secondary servers or resolvers' caches.

Default: propagation-delay + zone-max-ttl

rrsig-pre-refresh

A period how long at most before a signature refresh time the signature might be refreshed, in order to refresh RRSIGs in bigger batches on a frequently updated zone (avoid re-sign event too often).

Default: 1h (1 hour)

reproducible-signing

For ECDSA algorithms, generate RRSIG signatures deterministically (RFC 6979). Besides better theoretical cryptographic security, this mode allows significant speed-up of loading signed (by the same method) zones. However, the zone signing is a bit slower.

Default: off

nsec3

Specifies if NSEC3 will be used instead of NSEC.

Default: off

nsec3-iterations

A number of additional times the hashing is performed.

Default: 0

nsec3-opt-out

If set, NSEC3 records won't be created for insecure delegations. This speeds up the zone signing and reduces overall zone size.

Warning

NSEC3 with the Opt-Out bit set no longer works as a proof of non-existence in this zone.

Default: off

nsec3-salt-length

A length of a salt field in octets, which is appended to the original owner name before hashing.

Default: 8

nsec3-salt-lifetime

A validity period of newly issued salt field.

Zero value means infinity.

Special value -1 triggers re-salt every time when active ZSK changes. This optimizes the number of big changes to the zone.

Default: 30d (30 days)

signing-threads

When signing zone or update, use this number of threads for parallel signing.

Those are extra threads independent of Background workers.

Note

Some steps of the DNSSEC signing operation are not parallelized.

Default: 1 (no extra threads)

ksk-submission

A reference to submission section holding parameters of KSK submission checks.

Default: not set

ds-push

Optional references remote and remotes to authoritative DNS server of the parent's zone. The remote server must be configured to accept DS record updates via DDNS. Whenever a CDS record in the local zone is changed, the corresponding DS record is sent as a dynamic update (DDNS) to the parent DNS server. All previous DS records are deleted within the DDNS message. It's possible to manage both child and parent zones by the same Knot DNS server.

Note

This feature requires cds-cdnskey-publish not to be set to none.

Note

The mentioned change to CDS record usually means that a KSK roll-over is running and the new key being rolled-in is in "ready" state already for the period of propagation-delay.

Note

Module Onlinesign doesn't support DS push.

Default: not set

cds-cdnskey-publish

Controls if and how shall the CDS and CDNSKEY be published in the zone.

Possible values:

  • none – Never publish any CDS or CDNSKEY records in the zone.

  • delete-dnssec – Publish special CDS and CDNSKEY records indicating turning off DNSSEC.

  • rollover – Publish CDS and CDNSKEY records for ready and not yet active KSK (submission phase of KSK rollover).

  • always – Always publish one CDS and one CDNSKEY records for the current KSK.

  • double-ds – Always publish up to two CDS and two CDNSKEY records for ready and/or active KSKs.

Note

If the zone keys are managed manually, the CDS and CDNSKEY rrsets may contain more records depending on the keys available.

Warning

The double-ds value does not trigger double-DS roll-over method. That method is only supported when performed manually, with unset ksk-submission.

Default: rollover

cds-digest-type

Specify digest type for published CDS records.

Default: sha256

dnskey-management

Specify how the DNSKEY, CDNSKEY, and CDS RRSets at the zone apex are handled when (re-)signing the zone.

Possible values:

  • full – Upon every zone (re-)sign, delete all unknown DNSKEY, CDNSKEY, and CDS records and keep just those that are related to the zone keys stored in the KASP database.

  • incremental – Keep unknown DNSKEY, CDNSKEY, and CDS records in the zone, and modify server-managed records incrementally by employing changes in the KASP database.

Note

Prerequisites for incremental:

  • The Offline KSK isn't supported.

  • The delete-delay is long enough to cover possible daemon shutdown (e.g. due to server maintenance).

  • Avoided manual deletion of keys with keymgr.

Otherwise there might remain some DNSKEY records in the zone, belonging to deleted keys.

Default: full

offline-ksk

Specifies if Offline KSK feature is enabled.

Default: off

unsafe-operation

Turn off some DNSSEC safety features.

Possible values:

  • none – Nothing disabled.

  • no-check-keyset – Don't check active keys in present algorithms. This may lead to violation of RFC 4035#section-2.2.

  • no-update-dnskey – Don't maintain/update DNSKEY, CDNSKEY, and CDS records in the zone apex according to KASP database. Juste leave them as they are in the zone.

  • no-update-nsec – Don't maintain/update NSEC/NSEC3 chain. Leave all the records as they are in the zone.

  • no-update-expired – Don't update expired RRSIGs.

Multiple values may be specified.

Warning

This mode is intended for DNSSEC experts who understand the corresponding consequences.

Default: none

template section

A template is shareable zone settings, which can simplify configuration by reducing duplicates. A special default template (with the default identifier) can be used for global zone configuration or as an implicit configuration if a zone doesn't have another template specified.

template:
  - id: STR
    global-module: STR/STR ...
    # All zone options (excluding 'template' item)

Note

If an item is explicitly specified both in the referenced template and the zone, the template item value is overridden by the zone item value.

id

A template identifier.

global-module

An ordered list of references to query modules in the form of module_name or module_name/module_id. These modules apply to all queries.

Note

This option is only available in the default template.

Default: not set

zone section

Definition of zones served by the server.

zone:
  - domain: DNAME
    template: template_id
    storage: STR
    file: STR
    master: remote_id | remotes_id ...
    ddns-master: remote_id
    notify: remote_id | remotes_id ...
    acl: acl_id ...
    provide-ixfr: BOOL
    semantic-checks: BOOL | soft
    zonefile-sync: TIME
    zonefile-load: none | difference | difference-no-serial | whole
    journal-content: none | changes | all
    journal-max-usage: SIZE
    journal-max-depth: INT
    zone-max-size : SIZE
    adjust-threads: INT
    dnssec-signing: BOOL
    dnssec-validation: BOOL
    dnssec-policy: policy_id
    ds-push: remote_id | remotes_id ...
    zonemd-verify: BOOL
    zonemd-generate: none | zonemd-sha384 | zonemd-sha512 | remove
    serial-policy: increment | unixtime | dateserial
    refresh-min-interval: TIME
    refresh-max-interval: TIME
    retry-min-interval: TIME
    retry-max-interval: TIME
    expire-min-interval: TIME
    expire-max-interval: TIME
    catalog-role: none | interpret | generate | member
    catalog-template: template_id ...
    catalog-zone: DNAME
    catalog-group: STR
    module: STR/STR ...

domain

A zone name identifier.

template

A reference to a configuration template.

Default: not set or default (if the template exists)

storage

A data directory for storing zone files.

Default: ${localstatedir}/lib/knot (configured with --with-storage=path)

file

A path to the zone file. Non-absolute path (i.e. not starting with /) is relative to storage. It is also possible to use the following formatters:

  • %c[N] or %c[N-M] – Means the Nth character or a sequence of characters beginning from the Nth and ending with the Mth character of the textual zone name (see %s). The indexes are counted from 0 from the left. All dots (including the terminal one) are considered. If the character is not available, the formatter has no effect.

  • %l[N] – Means the Nth label of the textual zone name (see %s). The index is counted from 0 from the right (0 ~ TLD). If the label is not available, the formatter has no effect.

  • %s – Means the current zone name in the textual representation. The zone name doesn't include the terminating dot (the result for the root zone is the empty string!).

  • %% – Means the % character.

Warning

Beware of special characters which are escaped or encoded in the \DDD form where DDD is corresponding decimal ASCII code.

Default: storage/%s.zone

master

An ordered list of references remote and remotes to zone primary servers (formerly known as master servers).

Default: not set

ddns-master

A reference to zone primary master. If not specified, the first master server is used.

Default: not set

notify

An ordered list of references remote and remotes to secondary servers to which notify message is sent if the zone changes.

Default: not set

acl

An ordered list of references to ACL rules which can allow or disallow zone transfers, updates or incoming notifies.

Default: not set

provide-ixfr

If disabled, the server is forced to respond with AXFR to IXFR queries. If enabled, IXFR requests are responded normally.

Default: on

semantic-checks

Selects if extra zone semantic checks are used or impacts of the mandatory checks.

There are several mandatory checks which are always enabled and cannot be turned off. An error in a mandatory check causes the zone not to be loaded. Most of the mandatory checks can be weakened by setting soft, which allows the zone to be loaded even if the check fails.

If enabled, extra checks are used. These checks don't prevent the zone from loading.

The mandatory checks are applied to zone files, zone transfers, and updates via control interface. The extra checks are applied to zone files only!

Mandatory checks:

  • Missing SOA record at the zone apex (RFC 1034) (*)

  • An extra record exists together with a CNAME record except for RRSIG and NSEC (RFC 1034)

  • Multiple CNAME records with the same owner exist (RFC 1034)

  • DNAME record having a record under it (RFC 6672)

  • Multiple DNAME records with the same owner exist (RFC 6672)

  • NS record exists together with a DNAME record (RFC 6672)

(*) The marked check can't be weakened by the soft mode. All other mandatory checks are subject to the optional soft mode.

Extra checks:

  • Missing NS record at the zone apex

  • Missing glue A or AAAA record

  • Invalid DS or NSEC3PARAM record

  • CDS or CDNSKEY inconsistency

  • All other DNSSEC checks executed during dnssec-validation

Note

The soft mode allows the refresh event to ignore a CNAME response to a SOA query (malformed message) and triggers a zone bootstrap instead.

Default: off

zonefile-sync

The time after which the current zone in memory will be synced with a zone file on the disk (see file). The server will serve the latest zone even after a restart using zone journal, but the zone file on the disk will only be synced after zonefile-sync time has expired (or after manual zone flush). This is applicable when the zone is updated via IXFR, DDNS or automatic DNSSEC signing. In order to completely disable automatic zone file synchronization, set the value to -1. In that case, it is still possible to force a manual zone flush using the -f option.

Note

If you are serving large zones with frequent updates where the immediate sync with a zone file is not desirable, increase the value.

Default: 0 (immediate)

zonefile-load

Selects how the zone file contents are applied during zone load.

Possible values:

  • none – The zone file is not used at all.

  • difference – If the zone contents are already available during server start or reload, the difference is computed between them and the contents of the zone file. This difference is then checked for semantic errors and applied to the current zone contents.

  • difference-no-serial – Same as difference, but the SOA serial in the zone file is ignored, the server takes care of incrementing the serial automatically.

  • whole – Zone contents are loaded from the zone file.

When difference is configured and there are no zone contents yet (cold start and no zone contents in the journal), it behaves the same way as whole.

Default: whole

journal-content

Selects how the journal shall be used to store zone and its changes.

Possible values:

  • none – The journal is not used at all.

  • changes – Zone changes history is stored in journal.

  • all – Zone contents and history is stored in journal.

Default: changes

journal-max-usage

Policy how much space in journal DB will the zone's journal occupy.

Note

Journal DB may grow far above the sum of journal-max-usage across all zones, because of DB free space fragmentation.

Default: 100M (100 MiB)

journal-max-depth

Maximum history length of the journal.

Note

Zone-in-journal changeset isn't counted to the limit.

Minimum: 2

Default: 20

zone-max-size

Maximum size of the zone. The size is measured as size of the zone records in wire format without compression. The limit is enforced for incoming zone transfers and dynamic updates.

For incremental transfers (IXFR), the effective limit for the total size of the records in the transfer is twice the configured value. However the final size of the zone must satisfy the configured value.

Default: unlimited

adjust-threads

Parallelize internal zone adjusting procedures by using specified number of threads. This is useful with huge zones with NSEC3. Speedup observable at server startup and while processing NSEC3 re-salt.

Default: 1 (no extra threads)

dnssec-signing

If enabled, automatic DNSSEC signing for the zone is turned on.

Default: off

dnssec-validation

If enabled, the zone contents are validated for being correctly signed (including NSEC/NSEC3 chain) with DNSSEC signatures every time the zone is loaded or changed (including AXFR/IXFR).

When the validation fails, the zone being loaded or update being applied is cancelled with an error, and either none or previous zone state is published.

List of DNSSEC checks:

  • Every zone RRSet is correctly signed by at least one present DNSKEY.

  • DNSKEY RRSet is signed by KSK.

  • NSEC(3) RR exists for each name (unless opt-out) with correct bitmap.

  • Every NSEC(3) RR is linked to the lexicographically next one.

The validation is not affected by dnssec-policy configuration, except for signing-threads option, which specifies the number of threads for parallel validation.

Default: not set

Note

Redundant or garbage NSEC3 records are ignored.

This mode is not compatible with dnssec-signing.

dnssec-policy

A reference to DNSSEC signing policy.

Default: an imaginary policy with all default values

Note

A configured policy called "default" won't be used unless explicitly referenced.

ds-push

Per zone configuration of ds-push. This option overrides possible per policy option.

Default: not set

zonemd-verify

On each zone load/update, verify that ZONEMD is present in the zone and valid.

Note

Zone digest calculation may take much time and CPU on large zones.

Default: off

zonemd-generate

On each zone update, calculate ZONEMD and put it into the zone.

Possible values:

  • none – No action regarding ZONEMD.

  • zonemd-sha384 – Generate ZONEMD using SHA384 algorithm.

  • zonemd-sha512 – Generate ZONEMD using SHA512 algorithm.

  • remove – Remove any ZONEMD from the zone apex.

Default: none

serial-policy

Specifies how the zone serial is updated after a dynamic update or automatic DNSSEC signing. If the serial is changed by the dynamic update, no change is made.

Possible values:

  • increment – The serial is incremented according to serial number arithmetic.

  • unixtime – The serial is set to the current unix time.

  • dateserial – The 10-digit serial (YYYYMMDDnn) is incremented, the first 8 digits match the current iso-date.

Note

If the resulting serial for unixtime or dateserial is lower than or equal to the current serial (this happens e.g. when migrating from other policy or frequent updates), the serial is incremented instead.

To avoid user confusion, use dateserial only if you expect at most 100 updates per day per zone and unixtime only if you expect at most one update per second per zone.

Generated catalog zones use unixtime only.

Default: increment (unixtime for generated catalog zones)

refresh-min-interval

Forced minimum zone refresh interval (in seconds) to avoid flooding primary server.

Minimum: 2

Default: 2

refresh-max-interval

Forced maximum zone refresh interval (in seconds).

Default: not set

retry-min-interval

Forced minimum zone retry interval (in seconds) to avoid flooding primary server.

Minimum: 1

Default: 1

retry-max-interval

Forced maximum zone retry interval (in seconds).

Default: not set

expire-min-interval

Forced minimum zone expire interval (in seconds) to avoid flooding primary server.

Minimum: 3

Default: 3

expire-max-interval

Forced maximum zone expire interval (in seconds).

Default: not set

catalog-role

Trigger zone catalog feature. Possible values:

  • none – Not a catalog zone.

  • interpret – A catalog zone which is loaded from a zone file or XFR, and member zones shall be configured based on its contents.

  • generate – A catalog zone whose contents are generated according to assigned member zones.

  • member – A member zone that is assigned to one generated catalog zone.

Default: none

catalog-template

For the catalog member zones, the specified configuration template will be applied.

Multiple catalog templates may be defined. The first one is used unless the member zone has the group property defined, matching another catalog template.

Note

This option must be set if and only if catalog-role is interpret.

Nested catalog zones aren't supported. Therefore catalog templates can't use catalog-template, catalog-role, catalog-zone, and catalog-group options.

Default: not set

catalog-zone

Assign this member zone to specified generated catalog zone.

Note

This option must be set if and only if catalog-role is member.

The referenced catalog zone must exist and have catalog-role set to generate.

Default: not set

catalog-group

Assign this member zone to specified catalog group (configuration template).

Note

This option has effect if and only if catalog-role is member.

Default: not set

module

An ordered list of references to query modules in the form of module_name or module_name/module_id. These modules apply only to the current zone queries.

Default: not set