A full-featured SDK for Kong 2.3.x (with support 2.0.x, 2.1.x, and 2.2.x, and for migrating from 0.14.x, 1.1.x, 1.2.x, 1.4.x, and 1.5.x). Note that this is unofficial (meaning this project is in no way officially endorsed, recommended, or related to Kong as a company or an open-source project). It is also in no way related to the pet toy company by the same name (but hopefully that was obvious).
If you only plan on using the CLI tool, feel free to just use the Docker image:
docker pull jgnagy/skull_island
alias skull_island='docker run -it --rm -e KONG_ADMIN_URL=${KONG_ADMIN_URL:-http://host.docker.internal:8001} -u $(id -u) -v ${PWD}:${PWD} -w ${PWD} jgnagy/skull_island'
skull_island helpNote: Starting with 2.3.0, Skull Island only supports Ruby 3.0+. If you need to use Ruby 2.x, you'll need to use a gem version < 2.3.0. I'll maintain the 2.2.x line for Ruby 2.x compatibility for 2021.
To use the Skull Island gem, either:
gem install skull_islandOr add this to your Gemfile:
gem 'skull_island', '~> 2.3'Or add this to your .gemspec:
Gem::Specification.new do |spec|
# ...
spec.add_runtime_dependency 'skull_island', '~> 2.3'
# ...
endSkull Island comes with an executable called skull_island that leverages the SDK under the hood. Learn about what it can do via help:
$ skull_island help
Commands:
skull_island export [OPTIONS] [OUTPUT|-] # Export the current configuration to OUTPUT
skull_island help [COMMAND] # Describe available commands or one specific command
skull_island import [OPTIONS] [INPUT|-] # Import a configuration from INPUT
skull_island migrate [OPTIONS] [INPUT|-] [OUTPUT|-] # Migrate an older config from INPUT to OUTPUT
skull_island render [INPUT|-] # Render out preprocessed YAML for troubleshooting
skull_island reset # Fully reset a gateway (removing all config)
skull_island version # Display the current installed version of skull_island
Options:
[--verbose], [--no-verbose]To use the commands that interact with the Kong API, set environment variables for the required parameters:
KONG_ADMIN_URL='https://api-admin.mydomain.com' \
KONG_ADMIN_USERNAME='my-basicauth-user' \
KONG_ADMIN_PASSWORD='my-basicauth-password' \
skull_island ...Note that you can skip KONG_ADMIN_USERNAME and KONG_ADMIN_PASSWORD if you aren't using a basic-auth reverse-proxy in front of the Admin API.
Also note that if you're having SSL issues (such as with a private CA), you can have Ruby make use of a custom CA public key using SSL_CERT_FILE:
SSL_CERT_FILE=/path/to/cabundle.pem \
KONG_ADMIN_URL='https://api-admin.mydomain.com' \
KONG_ADMIN_USERNAME='my-basicauth-user' \
KONG_ADMIN_PASSWORD='my-basicauth-password' \
skull_island ...The CLI allows you to export an existing configuration to a YAML + ERB document (a YAML document with embedded Ruby). This format is helpful because it doesn't require you to know the IDs of resources, making your configuration portable.
The export command will default to outputting to STDOUT if you don't provide an output file location. Otherwise, simply specify the filename you'd like to export to:
KONG_ADMIN_URL='https://api-admin.mydomain.com' \
skull_island export /path/to/export.ymlYou can also get a little more information by turning on --verbose:
KONG_ADMIN_URL='https://api-admin.mydomain.com' \
skull_island export --verbose /path/to/export.ymlExporting, by default, exports the entire configuration of a Kong gateway, but will strip out special meta-data tags added by Skull Island to track projects. If, instead, you'd like to export only the configuration for a specific project, you can add --project foo (where foo is the name of your project) to export only those resources associated with it and maintain the special key in the exported YAML.
For most credential types, exporting works as expected (you'll see the plaintext value in the exported YAML). With BasicauthCredentials, however, this is not the case. This isn't a limitation of skull_island; rather, it is the expected behavior of the Admin API developers. This tool, when exporting these credentials, can only provide the salted-SHA1 hash of the password, and it does so by wrapping it in a special hash{} notation. This allows skull_island to distinguish between changing the value to the literal string and comparing the hashed values. The import process is also smart enough to compare plaintext to hashed values returned from the API for existing values, so it won't recreate credentials every run.
Skull Island also supports importing configurations (both partial and full) from a YAML + ERB document:
KONG_ADMIN_URL='https://api-admin.mydomain.com' \
skull_island import /path/to/export.ymlIt'll also read from STDIN if you don't specify a file path (or if you specify - as the path):
cat /path/to/export.yml | KONG_ADMIN_URL='https://api-admin.mydomain.com' skull_island import
# OR
KONG_ADMIN_URL='https://api-admin.mydomain.com' skull_island import < /path/to/export.ymlYou can also get a little more information by turning on --verbose:
KONG_ADMIN_URL='https://api-admin.mydomain.com' \
skull_island import --verbose /path/to/export.ymlImporting also supports a "dry run" functionality that shows you what it would do (but makes no changes) using --test:
KONG_ADMIN_URL='https://api-admin.mydomain.com' \
skull_island import --verbose --test /path/to/export.ymlNote that --test has a high likelihood of generating errors with a complicated import if required/dependent resources do not exist.
Skull Island 1.2.1 introduced the ability to use a special top-level key in the configuration called project that uses meta-data to track which resources belong to a project. This meta-data can safely be added at another time as this tool will "adopt" otherwise matching resources into a project.
To use this functionality, either add the project key to your configuration file (usually directly below the version key) with some value that will be unique on your gateway, or use --project foo (where foo is the name of your project) as a CLI flag.
When using the project feature of Skull Island, the CLI tool will automatically clean up old resources no longer found in your config file. This is, in fact, the only circumstance under which this tool actually removes resources. Use this feature with care, as it can delete large swaths of your configuration if used incorrectly. It is critical that this value is unique since this project functionality is used to delete resources.
If you're having trouble importing some YAML -- especially if you've done a lot of templating or have a lot of embedded ruby in your template -- you can use the render command to just output the raw, preprocessed YAML that Skull Island would attempt to import:
skull_island render /path/to/some/template.ymlThis is safe to do on your local machine since it doesn't connect to a gateway. It will leave lookup functions but otherwise will leave the YAML without embedded ruby. This can then be linted with any YAML linter to help discover YAML-related formatting issues, etc., that cause import issues.
With Skull Island, it is possible to migrate a configuration from a 0.14.x, 1.1.x, 1.2.x, 1.4.x, or 1.5.x gateway to the most recent compatible gateway. If you have a previous export, you can just run skull_island migrate /path/to/export.yml and you'll receive a 2.3 compatible config on standard out. If you'd prefer, you can have that config written to a file as well (just like the export command) like so:
skull_island migrate /path/to/export.yml /output/location/migrated.ymlWhile this hasn't been heavily tested for all possible use-cases, any configuration generated or usable by the '~> 0.14', '~> 1.2', or ~> 1.4 version of this gem should safely convert using the migration command. This tool also makes no guarantees about plugin functionality, configuration compatibility across versions, or that the same plugins are installed and available in your newer gateway. It should go without saying that you should test and confirm that all of your functionality was successfully migrated.
If you don't have a previous export, you'll need to install an older version of this gem using something like gem install --version '~> 0.14' skull_island, then perform an export, then you can switch back to the latest version of the gem for migrating and importing.
While it would be possible to make migration automatic for the import command, skull_island intentionally doesn't do this to avoid the appearance that the config is losslessly compatible across versions. In reality, the newer config version has additional features (like tagging) that are used heavily by skull_island. It makes sense to this author to maintain the migration component and the normal functionality as distinct features to encourage the use of the newer capabilities in 1.1 and beyond. That said, Skull Island does allow 1.x, 2.0.x, 2.1.x, and 2.2.x version configurations to be applied to 2.3 gateways, but not the opposite.
Skull Island can completely clear the configuration from a Kong instance using the reset command. THIS COMMAND WILL COMPLETELY CLEAR YOUR CONFIGURATION! Since this is a pretty serious command, it requires you to include --force, otherwise it simply exits with an error.
Fully resetting a gateway looks like this:
skull_island reset --forceYou can, of course, include --verbose to see skull_island do its work, though the output may be slightly misleading because of the cascading nature of deletions (e.g., deleting a Service will delete all Routes associated with it automatically).
You can also restrict the reset to just resources associated with a particular project using the --project flag:
skull_island reset --force --project fooThis assumes the project is called foo.
If you're wondering what version of skull_island is installed, use:
$ skull_island version
SkullIsland Version: 2.3.0The import/export/migrate CLI functions produce YAML with support for embedded Ruby (ERB). The file is structured like this (as an example):
---
version: '2.3'
project: FooV2
certificates: []
ca_certificates:
- cert: |-
-----BEGIN CERTIFICATE-----
MIIFUzCCA...
-----END CERTIFICATE-----
consumers:
- username: foo
custom_id: foo
acls:
- group: searchusers
credentials:
key-auth:
- key: q90r8908w09rqw9jfj09jq0f8y389
basic-auth:
- username: foo
password: bar
upstreams: []
services:
- name: apidocs
protocol: https
host: api.example.com
port: 443
path: "/v2/api-docs"
routes:
- paths:
- "/api-docs"
protocols:
- http
- https
strip_path: true
preserve_host: false
- name: search_api
retries: 5
protocol: https
host: api.example.com
port: 3737
connect_timeout: 30000
write_timeout: 30000
read_timeout: 30000
routes:
- methods:
- POST
paths:
- "/v2/search"
protocols:
- http
- https
regex_priority: 0
strip_path: true
preserve_host: false
plugins:
- name: key-auth
enabled: true
config:
anonymous: ''
hide_credentials: false
key_in_body: false
key_names:
- x-api-key
run_on_preflight: true
service: "<%= lookup :service, 'search_api' %>"
- name: acl
enabled: true
config:
hide_groups_header: false
whitelist:
- searchusers
service: "<%= lookup :service, 'search_api' %>"All top-level keys (other than version and project) require an Array as a parameter, either by providing a list of entries or an empty Array ([]), or they can be omitted entirely which is the same as providing an empty Array. The above shows how to use the lookup() function to refer to another resource. This "looks up" the resource type (service in this case) by name (search_api in this case) and resolves its id. This function can also be used to lookup a route or upstream by its name, or a consumer by its username. Note that Kong itself doesn't require route resources to have unique names, so you'll need to enforce that practice yourself for lookup to be useful for Routes. The lookup function also supports a third parameter to provide just the id value, rather than wrapping it in a key/value pair of id: 9992.... Simple call lookup like lookup :service, 'search_api', true to turn on this "raw id mode".
Note that while this configuration looks a lot like the DB-less configuration (and even may, at times, be interchangeable), this is merely a coincidence. Skull Island doesn't support the DB-less mode for Kong. This may potentially change in the future, but for now it is not a goal of this project.
While technically any Ruby is valid, the following are pretty helpful for templating your YAML files:
-
lookup(:service, 'foo')- This function resolves the ID of aservicenamedfoo. Lookup supports:consumer(looking up theusernameattribute),:service,:route, or:upstream(resolving thenameattribute). -
ENV.fetch('VARIABLE_NAME', 'default value')- This allows looking up the environment variableVARIABLE_NAMEand using its value, or, if it isn't defined, it usesdefault valueas the value. With this we could changehost: api.example.comtohost: <%= ENV.fetch('API_HOST', 'api.example.com') %>. With this, ifAPI_HOSTis provided, it'll use that, otherwise it will default toapi.example.com. This is especially helpful for sensitive information; you can version control the configuration but pass in things like credentials via environment variables at runtime.
Note also that 1.4.x and beyond of Skull Island support two phases of embedded ruby: first, a simple phase that treats the entire file as just text, allowing you to use the full power of ruby for things like loops, conditional logic, and more; the second phase is applied for individual attributes within the rendered YAML document. This is where the lookup() function above is used.
The API Client requires configuration before it can be used. For now, this is a matter of calling APIClient.configure(), passing a Hash, with Symbols for keys:
require 'skull_island'
include SkullIsland
APIClient.configure(
server: 'https://api-admin.mydomain.com',
username: 'my-basicauth-user',
password: 'my-basicauth-password'
)This assumes that a basic-auth reverse proxy sits in front of your Kong Admin API. If this isn't the case (it really should be), then just don't pass username and password and the API client will work just fine.
The API client provides a few helpful methods of its own. To learn about the overall service you've connected to (via the node information endpoint):
APIClient.about_service
# => {"plugins"=>...It is also possible to check on the server status of the node you're accessing via the node status endpoint:
APIClient.server_status
# => {"database"=>{"reachable"=>true...This SDK also makes use of automatic (and mostly unobtrusive) caching behind the scenes. As long as this tool is the only tool making changes to the Admin API (at least while it is being used), this should be fine. Eventually, there will be an option to disable this cache (at the cost of poor performance). For now, it is possible to query this cache and even flush it manually when required:
APIClient.lru_cache
# => #<SkullIsland::LRUCache:0x00007f9f1ebf3898 @max_size=1000...
APIClient.lru_cache.flush # this empties the cache and resets statistics
# => trueMost value provided by this SDK is through the ability to manipulate resources managed through the Admin API. These resources almost all have a few methods in common.
For example, finder methods like these:
# Get all instances of a resource type through `.all()`, returning a special collection class
Resources::Consumer.all
# => #<SkullIsland::ResourceCollection:0x00007f9f1e3f9b38...
# Find instances matching some criteria using `.where()`, returning the same kind of collection
Resources::Consumer.where(:username, /.*foo.*/) # finds all Consumers with a matching username
# => #<SkullIsland::ResourceCollection:0x00007f9f1e3351c0...
# Finding using other types of comparisons, like `<`
# Here, we find all Consumers made more than an hour ago
Resources::Consumer.where(
:created_at, (Time.now - 3600).to_datetime, comparison: :<
)
# => #<SkullIsland::ResourceCollection:0x00007f9f1e380924...
# Finder methods can also be chained
Resources::Consumer.where(:username, /.*foo.*/).and(:username, /\w{10,}/)
# => #<SkullIsland::ResourceCollection:0x00007f9f1e358964...
Resources::Consumer.where(:username, /.*foo.*/).or(:custom_id, /.*bar.*/)
# => #<SkullIsland::ResourceCollection:0x00007f9f1e568410...
# If you have the `id` of a particular resource, you can find it directly
Resources::Consumer.get('1cad3055-1027-459d-b76e-f590dc5f0071')
# => #<SkullIsland::Resources::Consumer:0x00007f9f201f6c58...Once you have a resource, you can modify it:
my_consumer = Resources::Consumer.get('1cad3055-1027-459d-b76e-f590dc5f0071')
my_consumer.username
# => "testuser"
my_consumer.tainted?
# => false
my_consumer.username = 'someuser'
my_consumer.tainted?
# => true
my_consumer.save
# => true
my_consumer.tainted?
# => false
my_consumer.username
# => "someuser"Some resource types are related to others, such as Routes and Services:
service = Resources::Service.all.first
# => #<SkullIsland::Resources::Services:0x00007f9f201f6f44...
service.routes
# => #<SkullIsland::ResourceCollection:0x00007f9f1e569e1d...
service.routes.size
# => 3
my_route = Resources::Route.new
my_route.hosts = ['example.com', 'example.org']
my_route.protocols = ['http', 'https']
my_route.strip_path = true
my_route.preserve_host = false
service.add_route!(my_route)
# => true
service.routes.size
# => 4From here, the SDK mostly wraps the attributes described in the Kong API Docs. For simplicity, I'll go over the resource types and attributes this SDK supports manipulating. Rely on the API documentation to determine which attributes are required and under which conditions.
These are used by the gateway to verify upstream certificates when connecting to them via HTTPS. These are not used to allow Kong to generate certificates. Thus, there is only a public key (cert) attribute for this resource.
resource = Resources::CACertificate.new
# These attributes can be set and read
resource.cert = '-----BEGIN CERTIFICATE-----...' # PEM-encoded public key
resource.tags = ['production', 'example'] # Array of tags
resource.save
# These attributes are read-only
resource.id
# => "1cad3055-1027-459d-b76e-f590dc5f0071"
resource.created_at
# => #<DateTime: 2018-07-17T12:51:28+00:00 ((2458317j,46288s,0n),+0s,2299161j)>Skull Island includes support for adding a virtual name attribute to CA Certificates (useful for maintaining them over time) based on a metadata tag that Skull Island will manage for you:
resource.name = 'internalca'
resource.name
# => "internalca"resource = Resources::Certificate.new
# These attributes can be set and read
resource.cert = '-----BEGIN CERTIFICATE-----...' # PEM-encoded public key
resource.key = '-----BEGIN RSA PRIVATE KEY-----...' # PEM-encoded private key
resource.snis = ['example.com', 'example.org'] # Array of names for which this cert is valid
resource.tags = ['production', 'example'] # Array of tags
resource.save
# These attributes are read-only
resource.id
# => "1cad3055-1027-459d-b76e-f590dc5f0071"
resource.created_at
# => #<DateTime: 2018-07-17T12:51:28+00:00 ((2458317j,46288s,0n),+0s,2299161j)>Skull Island includes support for adding a virtual name attribute to Certificates (useful for maintaining them over time) based on a metadata tag that Skull Island will manage for you:
resource.name = 'apicert'
resource.name
# => "apicert"Note that for Consumer credentials, only key-auth, jwt, and basic-auth are currently supported.
resource = Resources::Consumer.new
# These attributes can be set and read
resource.custom_id = 'user1' # A string
resource.username = 'user1' # A string
resource.tags = ['production', 'example'] # Array of tags
resource.save
# These attributes are read-only
resource.id
# => "1cad3055-1027-459d-b76e-f590dc5f0071"
resource.created_at
# => #<DateTime: 2018-07-17T12:51:28+00:00 ((2458317j,46288s,0n),+0s,2299161j)>
resource.plugins
# => #<SkullIsland::ResourceCollection:0x00007f9f1e564f3e...
resource.acls
# => #<SkullIsland::ResourceCollection:0x00007f9f1e765b3c...
resource.credentials
# => {}
resource.add_credential!(key: '932948e89e09e2989d8092') # adds a KeyauthCredential
# => true
resource.credentials['key-auth']
# => #<SkullIsland::ResourceCollection:0x00007f9f1e564f3f...
resource.credentials['key-auth'].first.key
# => "932948e89e09e2989d8092"
resource.add_credential!(username: 'test', password: 'passw0rd') # adds a BasicauthCredential
# => true
resource.credentials['basic-auth']
# => #<SkullIsland::ResourceCollection:0x00007f9f1e564f3f...
resource.credentials['basic-auth'].first.username
# => "test"
resource.add_acl!(group: 'somegroup')
# => trueNote that this doesn't install plugins; it only allows using them.
resource = Resources::Plugin.new
# These attributes can be set and read
resource.name = 'rate-limiting' # The name of the plugin
resource.enabled = true # A Boolean
resource.config = { 'minute' => 50, 'hour' => 1000 } # A Hash of config keys and values
resource.tags = ['production', 'example'] # Array of tags
# Either reference related resources by ID
resource.service = { 'id' => '5fd1z584-1adb-40a5-c042-63b19db49x21' }
resource.service
# => #<SkullIsland::Resources::Services:0x00007f9f201f6f44...
# Or reference related resources directly
resource.consumer = Resources::Consumer.get('a3dX2dh2-1adb-40a5-c042-63b19dbx83hF4')
resource.consumer
# => #<SkullIsland::Resources::Consumer:0x00007f9f201f6f98...
resource.route = Resources::Route.get('1cad3055-1027-459d-b76e-f590dc5f0023')
resource.route
# => #<SkullIsland::Resources::Route:0x00007f9f201f6f98...
resource.save
# These attributes are read-only
resource.id
# => "1cad3055-1027-459d-b76e-f590dc5f0071"
resource.created_at
# => #<DateTime: 2018-07-17T12:51:28+00:00 ((2458317j,46288s,0n),+0s,2299161j)>
# The resource class itself allows the following methods as well:
# This provides a list of plugin names that are enabled
Resources::Plugin.enabled_names
# => ["response-transformer",...
# This looks up the configuration schema for a particular plugin by its name
Resources::Plugin.schema('acl')
# => {"fields"=>{"hide_groups_header"=>{"default"=>false...resource = Resources::Route.new
# These attributes can be set and read
resource.name = 'example_route'
resource.hosts = ['example.com', 'example.org']
resource.protocols = ['https']
resource.methods = ['GET', 'POST']
resource.paths = ['/some/path']
resource.regex_priority = 10
resource.strip_path = false
resource.preserve_host = true
resource.tags = ['production', 'example']
# Or, for TCP/TLS routes
resource.protocols = ['tcp', 'tls']
resource.sources = [
{ "ip" => "10.1.0.0/16", "port" => 1234 }
]
resource.destinations = [
{ "ip" => "10.1.0.0/16", "port" => 1234 },
{ "ip" => "10.2.2.2" },
{ "port" => 9123 }
]
resource.snis = ['example.com', 'example.org']
# Either reference related resources by ID
resource.service = { 'id' => '4e13f54a-bbf1-47a8-8777-255fed7116f2' }
# Or reference related resources directly
resource.service = Resources::Service.get('4e13f54a-bbf1-47a8-8777-255fed7116f2')
resource.service
# => #<SkullIsland::Resources::Service:0x00007f9f201f6f98...
resource.save
# These attributes are read-only
resource.id
# => "1cad3055-1027-459d-b76e-f590dc5f0071"
resource.created_at
# => #<DateTime: 2018-07-17T12:51:28+00:00 ((2458317j,46288s,0n),+0s,2299161j)>
resource.updated_at
# => #<DateTime: 2018-07-17T12:51:28+00:00 ((2458317j,46288s,0n),+0s,2299161j)>
resource.plugins
# => #<SkullIsland::ResourceCollection:0x00007f9f1e564f3e...resource = Resources::Service.new
# These attributes can be set and read
resource.protocol = 'http'
resource.client_certificate = { 'id' => '77e32ff2-...' }
resource.connect_timeout = 60000
resource.host = 'example.com'
resource.port = 80
resource.path = '/api'
resource.name = 'example-service'
resource.retries = 10
resource.read_timeout = 60000
resource.write_timeout = 60000
resource.tags = ['production', 'example']
resource.save
# Add a related route
my_route = Resources::Route.get(...)
resource.add_route!(my_route) # adds a route to the service
# => true
# These attributes are read-only
resource.id
# => "1cad3055-1027-459d-b76e-f590dc5f0071"
resource.created_at
# => #<DateTime: 2018-07-17T12:51:28+00:00 ((2458317j,46288s,0n),+0s,2299161j)>
resource.updated_at
# => #<DateTime: 2018-07-17T12:51:28+00:00 ((2458317j,46288s,0n),+0s,2299161j)>
resource.routes
# => #<SkullIsland::ResourceCollection:0x00007f9f1e564f3f...
resource.plugins
# => #<SkullIsland::ResourceCollection:0x00007f9f1e564f3e...resource = Resources::Upstream.new
# These attributes can be set and read
resource.name = 'service.v1.xyz'
resource.algorithm = 'round-robin'
resource.hash_on = 'none'
resource.hash_fallback = 'none'
resource.slots = 1000
resource.healthchecks = {
'active' => {
'concurrency' => 5,
'healthy' => {
'http_statuses' => [200, 302],
'interval' => 0,
'successes' => 0
},
'http_path' => '/',
'timeout' => 1,
'unhealthy' => {
'http_failures' => 0,
'http_statuses' => [
429, 404, 500, 501, 502, 503, 504, 505
],
'interval' => 0,
'tcp_failures' => 0,
'timeouts' => 0
}
}
}
resource.tags = ['production', 'example']
resource.save
my_upstream_node = Resources::UpstreamTarget.new
my_upstream_node.target = '4.5.6.7:80'
my_upstream_node.weight = 15
resource.add_target!(my_upstream_node) # adds a target to the upstream
# => true
# These attributes are read-only
resource.id
# => "1cad3055-1027-459d-b76e-f590dc5f0071"
resource.created_at
# => #<DateTime: 2018-07-17T12:51:28+00:00 ((2458317j,46288s,0n),+0s,2299161j)>
resource.health # returns a Hash of all upstream targets and their health statuses
# => #<Hash...
resource.targets
# => #<SkullIsland::ResourceCollection:0x00007f9f1e564f3f...
resource.target('1bef3055-1027-459d-b76e-f590dc5f0071') # get an Upstream Target by idThe gem is available as open source under the terms of the MIT License.