Writing new DNS providers

Writing a new DNS provider is a relatively straightforward process. You essentially need to implement the providers.DNSServiceProvider interface. and the system takes care of the rest.

Please do note that if you submit a new provider you will be assigned bugs related to the provider in the future (unless you designate someone else as the maintainer). More details here.

Please follow the DNSControl Code Style Guide and the DNSControl Documentation Style Guide.

Overview

I'll ignore all the small stuff and get to the point.

A typical provider implements 3 methods and DNSControl takes care of the rest:

  • GetZoneRecords() -- Download the list of DNS records and return them as a list of RecordConfig structs.

  • GetZoneRecordsCorrections() -- Generate a list of corrections.

  • GetNameservers() -- Query the API and return the list of parent nameservers.

These three functions are all that's needed for dnscontrol preview and dnscontrol push.

The goal of GetZoneRecords() is to download all the DNS records, convert them to models.RecordConfig format, and return them as one big list (models.Records).

The goal of GetZoneRecordsCorrections() is to return a list of corrections. Each correction is a text string describing the change ("Delete CNAME record foo") and a function that, if called, will make the change (i.e. call the API and delete record foo). dnscontrol preview simply prints the text strings. dnscontrol push prints the strings and calls the functions. Because of how Go's functions work, the function will have everything it needs to make the change. Pretty cool, eh?

Calculating the difference between existing and desired is difficult. Luckily the work is done for you. GetZoneRecordsCorrections() calls a a function in the pkg/diff2 module that generates a list of changes (usually an ADD, CHANGE, or DELETE) that can easily be turned into the API calls mentioned previously.

So, what does all this mean?

It basically means that writing a provider is as simple as writing code that (1) downloads the existing records, (2) converts each records into models.RecordConfig, (3) write functions that perform adds, changes, and deletions.

If you are new to Go, there are plenty of providers you can copy from. In fact, many non-Go programmers have learned Go by contributing to DNSControl.

Now that you understand the general process, here are the details.

Step 1: General advice

A provider can be a DnsProvider, a Registrar, or both. We recommend you write the DnsProvider first, release it, and then write the Registrar if needed.

If you have any questions, please discuss them in the GitHub issue related to the request for this provider.

This document is constantly being updated. Please let us know what was confusing so we can update this document with advice for future authors (or even better send a PR!).

Step 2: Pick a base provider

It's a good idea to start by copying a similar provider.

How can you tell a provider is similar?

Each DNS provider's API falls into one of 4 category. Some update one DNS record at a time. Others, the only change they permit is to upload the entire zone even if only one record changed! Others are somewhere in between: all records at a label must be updated at once, or all records in a RecordSet (the label + rType).

In summary, provider APIs basically fall into four general categories:

  • Updates are done one record at a time (Record)

  • Updates are done one label at a time (Label)

  • Updates are done one label+type at a time (RecordSet)

  • Updates require the entire zone to be uploaded (Zone).

To determine your provider's category, review your API documentation.

To determine an existing provider's category, see which diff2.By*() function is used.

DNSControl provides 4 helper functions that do all the hard work for you. As input, they take the existing zone (what was downloaded via the API) and the desired zone (what is in dnsconfig.js). They return a list of instructions. Implement handlers for the instructions and DNSControl is able to perform dnscontrol push.

The functions are:

  • diff2.ByRecord() -- Updates are done one DNS record at a time. New records are added. Changes and deletes refer to an ID assigned to the record by the provider.

  • diff2.ByLabel() -- Updates are done for an entire label. Adds and changes are done by sending one or more records that will appear at that label (i.e. www.example.com). Deletes delete all records at that label.

  • diff2.ByRecordSet() -- Similar to ByLabel() but updates are done on the label+type level. If www.example.com has 2 A records and 2 MX records, updates must replace all the A records, or all the MX records, or add records of a different type.

  • diff2.ByZone() -- Updates are done by uploading the entire zone every time.

The file pkg/diff2/diff2.go has instructions about how to use the diff2 system.

Step 3: Create the driver skeleton

Create a directory for the provider called providers/name where name is all lowercase and represents the commonly-used name for the service.

The main driver should be called providers/name/nameProvider.go. The API abstraction is usually in a separate file (often called api.go).

Directory names should be consitent. It should be all lowercase and match the ALLCAPS provider name. Avoid _s.

Step 4: Activate the driver

Edit providers/_all/all.go. Add the provider list so DNSControl knows it exists.

Step 5: Implement

If you are implementing a DNS Service Provider:

Implement all the calls in the providers.DNSServiceProvider interface.

The function GetDomainCorrections() is a bit interesting. It returns a list of corrections to be made. These are in the form of functions that DNSControl can call to actually make the corrections.

If you are implementing a DNS Registrar:

Implement all the calls in the providers.Registrar interface.

The function GetRegistrarCorrections() returns a list of corrections to be made. These are in the form of functions that DNSControl can call to actually make the corrections.

Step 6: Unit Test

Make sure the existing unit tests work. Add unit tests for any complex algorithms in the new code.

Run the unit tests with this command:

go test ./...

Step 7: Integration Test

This is the most important kind of testing when adding a new provider. Integration tests use a test account and a test domain.

All records will be deleted from the test domain! Use a OTE domain or a real domain that isn't otherwise in use and can be destroyed.

Now you can run the integration tests.

For example, test BIND:

cd integrationTest              # NOTE: Not needed if already there
export BIND_DOMAIN='example.com'
go test -v -verbose -provider BIND

(BIND is a good place to start since it doesn't require API keys.)

This will run the tests on Amazon AWS Route53:

export R53_DOMAIN='dnscontroltest-r53.com'    # Use a test domain.
export R53_KEY_ID='CHANGE_TO_THE_ID'
export R53_KEY='CHANGE_TO_THE_KEY'
cd integrationTest              # NOTE: Not needed if already there
go test -v -verbose -provider ROUTE53

Some useful go test flags:

  • Run only certain tests using the -start and -end flags.

    • Rather than running all the tests, run just the tests you want.

    • These flags must be after the -provider FOO flag.

    • Example: go test -v -verbose -provider ROUTE53 -start 10 -end 20 run tests 10-20 inclusive.

    • Example: go test -v -verbose -provider ROUTE53 -start 5 -end 5 runs only test 5.

    • Example: go test -v -verbose -provider ROUTE53 -start 20 skip the first 19 tests.

    • Example: go test -v -verbose -provider ROUTE53 -end 20 only run the first 20 tests.

  • Slow tests? Add -timeout n to increase the timeout for tests

    • go test kills the tests after 10 minutes by default. Some providers need more time.

    • This flag must be before the -verbose flag. Usually it is the first flag after go test.

    • Example: go test -timeout 20m -v -verbose -provider CLOUDFLAREAPI

  • If a test will always fail because the provider doesn't support the feature, you can opt out of the test. Look at func makeTests() in integrationTest/integration_test.go for more details.

Step 8: Manual tests

This is optional.

There is a potential bug in how TXT records are handled. Sadly we haven't found an automated way to test for this bug. The manual steps are here in documentation/testing-txt-records.md

Step 9: Update docs, CICD and other files

  • Edit README.md:

    • Add the provider to the bullet list.

  • Edit .github/workflows/pr_test.yml

    • Add the name of the provider to the PROVIDERS list.

  • Edit documentation/providers.md:

  • Edit documentation/SUMMARY.md:

    • Add the provider to the "Providers" list.

  • Create documentation/provider/PROVIDERNAME.md:

    • Use one of the other files in that directory as a base.

  • Edit OWNERS:

    • Add the directory name and your GitHub username.

Need feedback? Submit a draft PR! It's a great way to get early feedback, ask about fixing a particular integration test, or request feedback.

Step 10: Capabilities

Some DNS providers have features that others do not. For example some support the SRV record. A provider announces what it can do using the capabilities system.

If a provider doesn't advertise a particular capability, the integration test system skips the appropriate tests. Therefore you might want to initially develop the provider with no particular capabilities advertised and code until all the integration tests work. Then enable capabilities one at a time to finish off the project.

Don't feel obligated to implement everything at once. In fact, we'd prefer a few small PRs than one big one. Focus on getting the basic provider working well before adding these extras.

Operational features have names like providers.CanUseSRV and providers.CanUseAlias. The list of optional "capabilities" are in the file dnscontrol/providers/providers.go (look for CanUseAlias).

Capabilities are processed early by DNSControl. For example if a provider doesn't support SRV records, DNSControl will error out when parsing dnscontrol.js rather than waiting until the API fails at the very end.

Enable optional capabilities in the nameProvider.go file and run the integration tests to see what works and what doesn't. Fix any bugs and repeat, repeat, repeat until you have all the capabilities you want to implement.

FYI: If a provider's capabilities changes, run go generate to update the documentation.

Step 11: Automated code tests

Run go vet and staticcheck and clean up any errors found.

go vet ./...
staticcheck ./...

Please use go vet from the newest release of Go.

golint is deprecated and frozen but it is still useful as it does a few checks that haven't been re-implemented in staticcheck. However golink fails on any file that uses generics, so be prepared to ignore errors about expected '(', found '[' (and 1 more errors)

How to install and run golint:

go get -u golang.org/x/lint/golint
go install golang.org/x/lint/golint
golint ./...

Step 12: Dependencies

See documentation/release-engineering.md for tips about managing modules and checking for outdated dependencies.

Step 13: Modify the release regexp

In the repo root, open .goreleaser.yml and add the provider to Provider-specific changes regexp.

Step 14: Check your work

These are the things we'll be checking when you submit the PR. Please try to complete all or as many of these as possible.

  1. Run go generate ./... to make sure all generated files are fresh.

  2. Make sure the following files were created and/or updated:

  • OWNERS

  • README.md

  • .github/workflows/pr_test.yml (The PROVIDERS list)

  • .goreleaser.yml (Search for Provider-specific changes)

  • documentation/SUMMARY.md

  • documentation/providers.md (the autogenerated table + the second one; make sure it is removed from the requested list)

  • documentation/provider/PROVIDERNAME.md

  • integrationTest/providers.json

    • providers/_all/all.go

  1. Review the code for style issues, remove debug statements, make sure all exported functions have a comment, and generally tighten up the code.

  2. Verify you're using the most recent version of anything you import. (See Step 12)

  3. Re-run the integration test one last time.

  • Post the results as a comment to your PR.

  1. Re-read the maintainer's responsibilities bullet list. By submitting a provider you agree to maintain it, respond to bugs, periodically re-run the integration test to verify nothing has broken, and if we don't hear from you for 2 months we may disable the provider.

Step 15: Submit a PR

At this point you can submit a PR.

Actually you can submit the PR even earlier if you just want feedback, input, or have questions. This is just a good stopping place to submit a PR if you haven't already.

Step 16: After the PR is merged

  1. Close any related GitHub issues.

  2. Would you like your provider to be tested automatically as part of every PR? Sure you would! Follow the instructions in Bring-Your-Own-Secrets for automated testing

Last updated