OctoDNS supports the following record types:
A
AAAA
ALIAS
CAA
CNAME
DNAME
LOC
MX
NAPTR
NS
PTR
SPF
SRV
SSHFP
TXT
URLFWD
Underlying provider support for each of these varies and some providers have extra requirements or limitations. In cases where a record type is not supported by a provider OctoDNS will ignore it there and continue to manage the record elsewhere. For example SSHFP
is supported by Dyn, but not Route53. If your source data includes an SSHFP record OctoDNS will keep it in sync on Dyn, but not consider it when evaluating the state of Route53. The best way to find out what types are supported by a provider is to look for its supports
method. If that method exists the logic will drive which records are supported and which are ignored. If the provider does not implement the method it will fall back to BaseProvider.supports
which indicates full support.
Adding new record types to OctoDNS is relatively straightforward, but will require careful evaluation of each provider to determine whether or not it will be supported and the addition of code in each to handle and test the new type.
- Dynamic Records - the preferred method for configuring geo-location, weights, and healthcheck based fallback between pools of services.
- Geo Records - the original implementation of geo-location based records, now superseded by Dynamic Records (above)
OctoDNS records and YamlProvider
's schema is essentially a 1:1 match. Properties on the objects will match keys in the config.
Each top-level key in the yaml file is a record name. Two common special cases are the root record ''
, and a wildcard '*'
.
---
'':
type: A
values:
- 1.2.3.4
- 1.2.3.5
'*':
type: CNAME
value: www.example.com.
www:
type: A
values:
- 1.2.3.4
- 1.2.3.5
www.sub:
type: A
values:
- 1.2.3.6
- 1.2.3.7
The above config lays out 4 records, A
s for example.com.
, www.example.com.
, and www.sub.example.com
and a wildcard CNAME
mapping *.example.com.
to www.example.com.
.
In the above example each name had a single record, but there are cases where a name will need to have multiple records associated with it. This can be accomplished by using a list.
---
'':
- type: A
values:
- 1.2.3.4
- 1.2.3.5
- type: MX
values:
- exchange: mx1.example.com.
preference: 10
- exchange: mx2.example.com.
preference: 10
Each record type has a corresponding set of required data. The easiest way to determine what's required is probably to look at the record object in octodns/record/__init__.py
. You may also utilize octodns-validate
which will throw errors about what's missing when run.
type
is required for all records. ttl
is optional. When TTL is not specified the YamlProvider
's default will be used. In any situation where an array of values
can be used you can opt to go with value
as a single item if there's only one.
octoDNS is fairly strict in terms of standards compliance and is opinionated in terms of best practices. Examples of former include SRV record naming requirements and the latter that ALIAS records are constrained to the root of zones. The strictness and support of providers varies so you may encounter existing records that fail validation when you try to dump them or you may even have use cases for which you need to create or preserve records that don't validate. octoDNS's solution to this is the lenient
flag.
It's best to think of the lenient
flag as "I know what I'm doing and accept any problems I run across." The main reason being is that some providers may allow the non-compliant setup and others may not. The behavior of the non-compliant records may even vary from one provider to another. Caveat emptor.
If you're trying to import a zone into octoDNS config file using octodns-dump
which fails due to validation errors you can supply the --lenient
argument to tell octoDNS that you acknowledge that things aren't lining up with its expectations, but you'd like it to go ahead anyway. This will do its best to populate the zone and dump the results out into an octoDNS zone file and include the non-compliant bits. If you go to use that config file octoDNS will again complain about the validation problems. You can correct them in cases where that makes sense, but if you need to preserve the non-compliant records read on for options.
When there are non-compliant records configured in Yaml you can add the following to tell octoDNS to do it's best to proceed with them anyway. If you use --lenient
above to dump a zone and you'd like to sync it as-is you can mark the problematic records this way.
'not-root':
octodns:
lenient: true
type: ALIAS
values: something.else.com.
If you'd like to enable lenience for a whole zone you can do so with the following, thought it's strongly encouraged to mark things at record level when possible. The most common case where things may need to be done at the zone level is when using something other than YamlProvider
as a source, e.g. syncing from Route53Provider
to Ns1Provider
when there are non-compliant records in the zone in Route53.
non-compliant-zone.com.:
lenient: true
sources:
- route53
targets:
- ns1
OctoDNS currently provides the ability to limit the number of updates/deletes on DNS records by configuring a percentage of allowed operations as a threshold. If left unconfigured, suitable defaults take over instead. In the below example, the Dyn provider is configured with limits of 40% on both update and delete operations over all the records present.
dyn:
class: octodns.provider.dyn.DynProvider
update_pcent_threshold: 0.4
delete_pcent_threshold: 0.4