Commit f9c59b77 authored by Daniel Salzman's avatar Daniel Salzman

doc: move modules into separate module directories

parent 83b56d9a
......@@ -457,6 +457,40 @@ of the limitations will be hopefully removed in the near future.
- Legacy key export is not implemented.
- DS record export is not implemented.
.. _query-modules:
Query modules
Knot DNS supports configurable query modules that can alter the way
queries are processed. Each query requires a finite number of steps to
be resolved. We call this set of steps a *query plan*, an abstraction
that groups these steps into several stages.
* Before-query processing
* Answer, Authority, Additional records packet sections processing
* After-query processing
For example, processing an Internet-class query needs to find an
answer. Then based on the previous state, it may also append an
authority SOA or provide additional records. Each of these actions
represents a 'processing step'. Now, if a query module is loaded for a
zone, it is provided with an implicit query plan which can be extended
by the module or even changed altogether.
A module is active if its name, which includes the ``mod-`` prefix, is assigned
to the zone/template :ref:`zone_module` option or to the *default* template
:ref:`template_global-module` option if activating for all queries.
If the module is configurable, a corresponding module section with
an identifier must be created and then referenced in the form of
``module_name/module_id``. See :ref:`Modules` for the list of available modules.
.. NOTE::
Query modules are processed in the order they are specified in the
zone/template configuration. In most cases, the recommended order is::
mod-synthrecord, mod-onlinesign, mod-rrl, mod-dnstap, mod-stats
Performance Tuning
......@@ -10,10 +10,10 @@ Welcome to Knot DNS's documentation!
This diff is collapsed.
This diff is collapsed.
This diff is collapsed.
.. _mod-dnsproxy:
``dnsproxy`` – Tiny DNS proxy
The module forwards all queries, or all specific zone queries if configured
per zone, to the indicated server for resolution. If configured in the fallback
mode, only localy unsatisfied queries are forwarded. I.e. a tiny DNS proxy.
There are several uses of this feature:
* A substitute public-facing server in front of the real one
* Local zones (poor man's "views"), rest is forwarded to the public-facing server
* etc.
.. NOTE::
The module does not alter the query/response as the resolver would,
and the original transport protocol is kept as well.
The configuration is straightforward and just a single remote server is
- id: hidden
- id: default
remote: hidden
fallback: on
- id: default
global-module: mod-dnsproxy/default
- domain:
When clients query for anything in the ````, they will be
responded to locally. The rest of the requests will be forwarded to the
specified server (```` in this case).
Module reference
- id: STR
remote: remote_id
timeout: INT
fallback: BOOL
catch-nxdomain: BOOL
.. _mod-dnsproxy_id:
A module identifier.
.. _mod-dnsproxy_remote:
A :ref:`reference<remote_id>` to a remote server where the queries are
forwarded to.
.. _mod-dnsproxy_timeout:
A remote response timeout in milliseconds.
*Default:* 500
.. _mod-dnsproxy_fallback:
If enabled, localy unsatisfied queries leading to REFUSED (no zone) are forwarded.
If disabled, all queries are directly forwarded without any local attempts
to resolve them.
*Default:* on
.. _mod-dnsproxy_catch-nxdomain:
If enabled, localy unsatisfied queries leading to NXDOMAIN are forwarded.
This option is only relevant in the fallback mode.
*Default:* off
.. _mod-dnstap:
``dnstap`` – Dnstap traffic logging
A module for query and response logging based on the dnstap_ library.
You can capture either all or zone-specific queries and responses; usually
you want to do the former.
The configuration comprises only a :ref:`mod-dnstap_sink` path parameter,
which can be either a file or a UNIX socket::
- id: capture_all
sink: /tmp/capture.tap
- id: default
global-module: mod-dnstap/capture_all
.. NOTE::
To be able to use a Unix socket you need an external program to create it.
Knot DNS connects to it as a client using the libfstrm library. It operates
exactly like syslog. See `here
<>`_ for
more details.
.. NOTE::
Dnstap log files can also be created or read using ``kdig``.
.. _dnstap:
Module reference
For all queries logging, use this module in the *default* template. For
zone-specific logging, use this module in the proper zone configuration.
- id: STR
sink: STR
identity: STR
version: STR
log-queries: BOOL
log-responses: BOOL
.. _mod-dnstap_id:
A module identifier.
.. _mod-dnstap_sink:
A sink path, which can be either a file or a UNIX socket when prefixed with
.. _mod-dnstap_identity:
A DNS server identity. Set empty value to disable.
*Default:* FQDN hostname
.. _mod-dnstap_version:
A DNS server version. Set empty value to disable.
*Default:* server version
.. _mod-dnstap_log-queries:
If enabled, query messages will be logged.
*Default:* on
.. _mod-dnstap_log-responses:
If enabled, response messages will be logged.
*Default:* on
.. _mod-noudp:
``noudp`` — No UDP response
The module sends empty truncated response to any UDP query. TCP queries are
not affected.
To enable this module globally, you need to add something like the following
to the configuration file::
- id: default
global-module: mod-noudp
.. NOTE::
This module is not configurable.
.. _mod-onlinesign:
``onlinesign`` — Online DNSSEC signing
The module provides online DNSSEC signing. Instead of pre-computing the zone
signatures when the zone is loaded into the server or instead of loading an
externally signed zone, the signatures are computed on-the-fly during
The main purpose of the module is to enable authenticated responses with
zones which use other dynamic module (e.g., automatic reverse record
synthesis) because these zones cannot be pre-signed. However, it can be also
used as a simple signing solution for zones with low traffic and also as
a protection against zone content enumeration (zone walking).
In order to minimize the number of computed signatures per query, the module
produces a bit different responses from the responses that would be sent if
the zone was pre-signed. Still, the responses should be perfectly valid for
a DNSSEC validating resolver.
Differences from statically signed zones:
* The NSEC records are constructed as Minimally Covering NSEC Records
(see Appendix A in :rfc:`7129`). Therefore the generated domain names cover
the complete domain name space in the zone's authority.
* NXDOMAIN responses are promoted to NODATA responses. The module proves
that the query type does not exist rather than that the domain name does not
* Domain names matching a wildcard are expanded. The module pretends and proves
that the domain name exists rather than proving a presence of the wildcard.
Records synthesized by the module:
* DNSKEY record is synthesized in the zone apex and includes public key
material for the active signing key.
* NSEC records are synthesized as needed.
* RRSIG records are synthesized for authoritative content of the zone.
Known issues:
* The delegations are not signed correctly.
* Some CNAME records are not signed correctly.
* The automatic policy-based key rotation does not work. The rotation events are
invoked just at server (re)load.
* Online-sign module always enforces Single-Type Signing scheme.
* Only one active signing key can be used.
* Key rollover is not possible.
* The NSEC records may differ for one domain name if queried for different
types. This is an implementation shortcoming as the dynamic modules
cooperate loosely. Possible synthesis of a type by other module cannot
be predicted. This dissimilarity should not affect response validation,
even with validators performing `aggressive negative caching
* The NSEC proofs will work well with other dynamic modules only if the
modules synthesize only A and AAAA records. If synthesis of other type
is required, please, report this information to Knot DNS developers.
* Enable the module in the zone configuration with the default signing policy::
- domain:
module: mod-onlinesign
Or with an explicit signing policy::
- id: rsa
algorithm: RSASHA256
zsk-size: 2048
- id: explicit
policy: rsa
- domain:
module: mod-onlinesign/explicit
Or use manual policy in an analogous manner, see
:ref:`Manual key management<dnssec-manual-key-management>`.
.. NOTE::
Only id, manual, keystore, algorithm, zsk-size, and rrsig-lifetime policy items are
relevant to this module. If no rrsig-lifetime is configured, the
default value is 25 hours.
* Make sure the zone is not signed and also that the automatic signing is
disabled. All is set, you are good to go. Reload (or start) the server:
.. code-block:: console
$ knotc reload
The following example stacks the online signing with reverse record synthesis
- id: lan-forward
type: forward
prefix: ip-
ttl: 1200
- domain:
module: [mod-synthrecord/lan-forward, mod-onlinesign]
Module reference
- id: STR
policy: STR
.. _mod-onlinesign_id:
A module identifier.
.. _mod-onlinesign_policy:
A :ref:`reference<policy_id>` to DNSSEC signing policy. A special *default*
value can be used for the default policy settings.
.. _mod-rosedb:
``rosedb`` – Static resource records
The module provides a mean to override responses for certain queries before
the record is searched in the available zones. The module comes with the
``rosedb_tool`` tool used to manipulate the database of static records.
For example, let's suppose we have a database of following records:
.. code-block:: none 3600 IN A 3600 IN A 3600 IN AAAA ::1
And we query the nameserver with the following:
.. code-block:: console
$ kdig IN A
... returns NOERROR,
$ kdig IN A
... returns NOERROR,
$ kdig IN A
... returns NOERROR,
$ kdig IN AAAA
... returns NOERROR, NODATA
$ kdig IN AAAA
... returns NOERROR, ::1
An entry in the database matches anything at the same or a lower domain
level, i.e. '' matches '' as well.
This can be utilized to create catch-all entries.
You can also add authority information for the entries, provided you create
SOA + NS records for a name, like so:
.. code-block:: none 3600 IN SOA master host 1 3600 60 3600 3600 3600 IN NS 3600 IN NS 3600 IN A 3600 IN A
In this case, the responses will:
1. Be authoritative (AA flag set)
2. Provide an authority section (SOA + NS)
3. Be NXDOMAIN if the name is found *(i.e. the 'IN AAAA' from
the example)*, but not the RR type *(this is to allow the synthesis of
negative responses)*
The SOA record applies only to the '', not to any other
record (not even those of its subdomains). From this point of view, all records
in the database are unrelated and not hierarchical. The idea is to provide
subtree isolation for each entry.
In addition, the module is able to log matching queries via remote syslog if
you specify a syslog address endpoint and an optional string code.
* Create the entries in the database:
.. code-block:: console
$ mkdir /tmp/static_rrdb
$ # No logging
$ rosedb_tool /tmp/static_rrdb add A 3600 "" "-" "-"
$ # Logging as 'www_query' to Syslog at
$ rosedb_tool /tmp/static_rrdb add A 3600 "" \
"www_query" ""
$ # Logging as 'ipv6_query' to Syslog at
$ rosedb_tool /tmp/static_rrdb add AAAA 3600 "::1" \
"ipv6_query" ""
$ # Verify settings
$ rosedb_tool /tmp/static_rrdb list A RDATA=10B www_query AAAA RDATA=22B ipv6_query A RDATA=10B - -
.. NOTE::
The database may be modified later on while the server is running.
* Configure the query module::
- id: default
dbdir: /tmp/static_rrdb
- id: default
global-module: mod-rosedb/default
The module accepts just one parameter – the path to the directory where
the database will be stored.
* Start the server:
.. code-block:: console
$ knotd -c knot.conf
* Verify the running instance:
.. code-block:: console
$ kdig @ A
Module reference
- id: STR
dbdir: STR
.. _mod-rosedb_id:
A module identifier.
.. _mod-rosedb_dbdir:
A path to the directory where the database is stored.
.. _mod-rrl:
``rrl`` — Response rate limiting
Response rate limiting (RRL) is a method to combat DNS reflection amplification
attacks. These attacks rely on the fact that source address of a UDP query
can be forged, and without a worldwide deployment of `BCP38
<>`_, such a forgery cannot be prevented.
An attacker can use a DNS server (or multiple servers) as an amplification
source and can flood a victim with a large number of unsolicited DNS responses.
The RRL lowers the amplification factor of these attacks by sending some of
the responses as truncated or by dropping them altogether.
.. NOTE::
The module introduces two statistics counters. The number of slipped and
dropped responses.
You can enable RRL by setting the module globally or per zone.
- id: default
rate-limit: 200 # Allow 200 resp/s for each flow
slip: 2 # Every other response slips
- id: default
global-module: mod-rrl/default # Enable RRL globally
Module reference
- id: STR
rate-limit: INT
slip: INT
table-size: INT
whitelist: ADDR[/INT] | ADDR-ADDR ...
.. _mod-rrl_id:
A module identifier.
.. _mod-rrl_rate-limit:
Rate limiting is based on the token bucket scheme. A rate basically
represents a number of tokens available each second. Each response is
processed and classified (based on several discriminators, e.g.
source netblock, query type, zone name, rcode, etc.). Classified responses are
then hashed and assigned to a bucket containing number of available
tokens, timestamp and metadata. When available tokens are exhausted,
response is dropped or sent as truncated (see :ref:`mod-rrl_slip`).
Number of available tokens is recalculated each second.
.. _mod-rrl_table-size:
Size of the hash table in a number of buckets. The larger the hash table, the lesser
the probability of a hash collision, but at the expense of additional memory costs.
Each bucket is estimated roughly to 32 bytes. The size should be selected as
a reasonably large prime due to better hash function distribution properties.
Hash table is internally chained and works well up to a fill rate of 90 %, general
rule of thumb is to select a prime near 1.2 * maximum_qps.
*Default:* 393241
.. _mod-rrl_slip:
As attacks using DNS/UDP are usually based on a forged source address,
an attacker could deny services to the victim's netblock if all
responses would be completely blocked. The idea behind SLIP mechanism
is to send each N\ :sup:`th` response as truncated, thus allowing client to
reconnect via TCP for at least some degree of service. It is worth
noting, that some responses can't be truncated (e.g. SERVFAIL).
- Setting the value to **0** will cause that all rate-limited responses will
be dropped. The outbound bandwidth and packet rate will be strictly capped
by the :ref:`mod-rrl_rate-limit` option. All legitimate requestors affected
by the limit will face denial of service and will observe excessive timeouts.
Therefore this setting is not recommended.
- Setting the value to **1** will cause that all rate-limited responses will
be sent as truncated. The amplification factor of the attack will be reduced,
but the outbound data bandwidth won't be lower than the incoming bandwidth.
Also the outbound packet rate will be the same as without RRL.
- Setting the value to **2** will cause that half of the rate-limited responses
will be dropped, the other half will be sent as truncated. With this
configuration, both outbound bandwidth and packet rate will be lower than the
inbound. On the other hand, the dropped responses enlarge the time window
for possible cache poisoning attack on the resolver.
- Setting the value to anything **larger than 2** will keep on decreasing
the outgoing rate-limited bandwidth, packet rate, and chances to notify
legitimate requestors to reconnect using TCP. These attributes are inversely
proportional to the configured value. Setting the value high is not advisable.
*Default:* 1
.. _mod-rrl_whitelist:
A list of IP addresses, network subnets, or network ranges to exempt from
rate limiting. Empty list means that no incoming connection will be
*Default:* not set
.. _mod-stats:
``stats`` — Query statistics
The module extends server statistics with incoming DNS request and corresponding
response counters, such as used network protocol, total number of responded bytes,
etc (see module reference for full list of supported counters).
This module should be configured as the last module.
.. NOTE::
Server initiated communication (outgoing NOTIFY, incoming \*XFR,...) is not
counted by this module.
.. NOTE::
Leading 16-bit message size over TCP is not considered.
Common statistics with default module configuration::
- id: default
global-module: mod-stats
Per zone statistics with explicit module configuration::
- id: custom
edns-presence: on
query-type: on
- id: default
module: mod-stats/custom
Module reference
- id: STR
request-protocol: BOOL
server-operation: BOOL
request-bytes: BOOL
response-bytes: BOOL
edns-presence: BOOL
flag-presence: BOOL
response-code: BOOL
reply-nodata: BOOL
query-type: BOOL
query-size: BOOL
reply-size: BOOL
.. _mod-stats_id:
A module identifier.
.. _mod-stats_request-protocol:
If enabled, all incoming requests are counted by the network protocol: