@arobson/hikaru
Advanced tools
A continuous deployment service and integration helper for kubernetes
A deployment and continuous delivery tool for kubernetes. 100 internets if you get the reference. 1000 internets if you get why.
hikaru was built specifically to make it easy to deploy and maintain kubernetes clusters.
hikaru assumes that git and Docker interaction that require credentials will have them based on the context the commands are run in.
hikaru does not accept git credentials nor will it store and fetch them. If the cluster spec provided by a call isn't public, then the expectation is that the git client already has the credentials necessary to access the private repository.
hikaru also does not accept docker credentials nor will it store and fetch them. When an image is specified by a manifest or by an upgrade call that is not publicly available, the expectation is that the Kubernetes cluster has been configured with credentials necessary to acquire the Docker image.
hikaru requires mcgonagall style cluster specifications. You'll get the most mileage out of it by adopting Dockyard to build your Docker images or at least using a compatible tagging approach (see: buildgoggles). These complimentary aspects come largely from how hikaru infers metadata about Kubernetes resources based on the Docker image name and tag.
hikaru provides 3 possible models for interaction:
hikaru's upgradeImage call (designed to call via Web Hook) determines which Deployments would qualify for an upgrade based on the image's tag and performs rolling upgrades. Limitations to this behavior can be introduced via metadata on the Kubernetes manifests themselves.
In practice, hikaru is able to perform automated rolling upgrades out of the box with no additional metadata on the manifests. Metadata added to the manifests provides ways to control the criteria hikaru uses to perform upgrades.
By default, hikaru will upgrade images if the version or build is newer and the following criteria match:
imageNameimageOwnerownerrepobranchThe following labels act like additional filters that hikaru will use against the image metadata it determines from the tag. If these labels are present under the manifest's spec.template.metadata.labels then they will be used to determine whether a new image is compatible with the manifest.
This allows you to filter out builds from owners, branches, etc. that you don't want hikaru to deploy over a certain environment/manifest:
owner - limit deployments to a specific repo ownerrepo - limit deployments to a specific repobranch - limit deployments to a specific branchversion - limit to new builds of a specific versioncommit - !caution! locks the deployment to a specific commitfilter - provide a comma delimited list of fields to filter on based on the current imageThe filter label was introduced so that you can dynamically assign filter values based on the current image without having to specify them in the manifest explicitly.
This allows you to change what's in effect by forcing an upgrade to the image but not having to change the labels themselves.
Examples:
Lock a resource down to the same owner, repository, and branch:
Note: code snippets demonstrate mcgonagall spec
labels = 'filter=owner,repo,branch'
Lock a resource down to the same owner, repository, branch, and version:
labels = 'owner,repo,branch,version'
Possible values:
imageNameimageOwnerownerrepobranchfullVersionversionbuildcommithikaru understands how to process the following tag styles:
latest - the latest build from our master branchmajor - the major part of the semantic versionmajor.minor - the major and minor parts of the semantic versionmajor.minor.patch - full semantic versionversion-prebuild - any previous semantic version example followed by a prebuild specifier{version}_{build}_{commit-slug}{branch}_{version}_{build}_{commit-slug}{owner}_{repo}_{branch}_{version}_{build}_{commit-slug} to allow for builds of forks and builds where multiple images are generated from a single repo.hikaru will attempt to filter out additional tag noise and also treat any -{label} following a version number as a pre-release.
hikaru will determine or infer 6 pieces of information about a Docker image:
redis - image = redis/redismhart/alpine-node - image = mhart/alpine-nodeofficial image
npm/app:arobson_master_1.0.1_10_abcd1234 - owner = arobsonnpm/app:latest - owner = npmnginx:13-alpine - owner = officialmasternpm/app:arobson_test_1.0.1_10_abcd1234 - branch = testnpm/app:now-with-more-stuff_1.0.1_10_abcd1234 - branch = now-with-more-stuffnpm/app:1.0 - branch = masternpm/app:1 - branch = masternpm/app:latest - branch = masternpm/app:latest - latestnpm/app:1 - latest 1.x.xnpm/app:1.1 - latest 1.1.xnpm/app:arobson_test_1.0.1_10_abcd1234 - 1.0.1lock when testing development buildsHere are a few examples to demonstrate how hikaru would read a few tags.
nginx
{
imageOwner: 'official', // inferred
imageName: 'nginx',
owner: 'official', // inferred
repo: 'nginx', // inferred
branch: 'master', // inferred
version: 'latest', // inferred
build: undefined,
commit: undefined
}
arobson/hikaru:1
{
imageOwner: 'arobson',
imageName: 'hikaru',
owner: 'arobson', // inferred
repo: 'hikaru', // inferred
branch: 'master', // inferred
version: '1.x.x',
build: undefined,
commit: undefined
}
arobson/kickerd:1.1-alpha_demo_10
{
imageOwner: 'arobson',
imageName: 'kickerd',
owner: 'arobson', // inferred
repo: 'kickerd', // inferred
branch: 'master', // inferred
version: '1.1.x-alpha',
build: '10',
commit: undefined
}
arobson/hikaru:dev_1.1.0_1_a1b2c3d4
{
imageOwner: 'arobson',
imageName: 'hikaru',
owner: 'arobson', // inferred
repo: 'hikaru', // inferred
branch: 'dev',
version: '1.1.0',
build: '1',
commit: 'a1b2c3d4'
}
npm/awesome:arobson_secret_master_0.0.1_10_a00b00c2
{
imageOwner: 'npm',
imageName: 'awesome',
owner: 'arobson',
repo: 'secret',
branch: 'master',
version: '0.0.1',
build: '10',
commit: 'a00b00c2'
}
Deploy From Git
POST /api/cluster/{gitHost}/{repoOwner}/{repoName}
POST /api/cluster/{gitHost}/{repoOwner}/{repoName}/{branch}
application/json
The POST body allows you to supply any tokens that exist in the spec. If tokens are present in the spec and not provided, a 400 will result with a list of the missing tokens.
Deploy From Tarball
POST /api/cluster
application/x-tar
The POST body will have to be multi-part to include both the tarball and a set of tokens if the specification requires them.
The delete actions requires the same specification used to create it (including any tokens). That's because hikaru reverses the deploy steps rather than just aggressively deleting everything from the cluster.
This also prevents you from deleting resources that weren't deployed as part of your spec.
Remove From Git
DELETE /api/cluster/{gitHost}/{repoOwner}/{repoName}
DELETE /api/cluster/{gitHost}/{repoOwner}/{repoName}/{branch}
application/json
The DELETE body allows you to supply any tokens that exist in the spec. If tokens are present in the spec and not provided, a 400 will result with a list of the missing tokens.
Remove From Tarball
DELETE /api/cluster
application/x-tar
The DELETE body will have to be multi-part to include both the tarball and a set of tokens if the specification requires them.
Returns a hash containing lists of resources.
upgrade has the list of resources eligible for upgrade.obsolete is the list of compatible resources that have a newer version than the posted imageequal is the list of compatible resources that already have the imageerror is the list of resources that were ignored which includes a diff property with a brief explanation of why they were ignoredGET /api/image/{image}?filter=
GET /api/image/{repo}/{image}?filter=
GET /api/image/{registry}/{repo}/{image}?filter=
The filter query parameters accepts a comma delimited list of fields that you want used to determine upgrade eligibility. Valid fields are:
imageNameimageOwnerownerrepobranchfullVersionversionbuildcommitThe reason for the multiple forms may not be obvious until you see examples:
GET /api/image/nginx:1.13-alpine
GET /api/image/arobson/hikaru:latest
GET /api/image/quay.io/coreos/etcd:v3.3.3
You could make the last form more permissive by telling it to only consider the imageOwner:
GET /api/image/quay.io/coreos/etcd:v3.3.3?filter=imageOwner
So that it would upgrade any resource using any etcd image regardless of whether or not it was the coreos Docker image or not.
Returns a hash containing lists of resources.
upgrade has the list of resources upgraded.obsolete is the list of compatible resources that have a newer version than the posted imageequal is the list of compatible resources that already have the imageerror is the list of resources that were ignored which includes a diff property with a brief explanation of why they were ignoredPOST /api/image/{image}?filter=
POST /api/image/{repo}/{image}?filter=
POST /api/image/{registry}/{repo}/{image}?filter=
The filter query parameters accepts a comma delimited list of fields that you want used to determine upgrade eligibility. Valid fields are:
imageNameimageOwnerownerrepobranchfullVersionversionbuildcommitThe reason for the multiple forms may not be obvious until you see examples:
POST /api/image/nginx:1.13-alpine
POST /api/image/arobson/hikaru:latest
POST /api/image/quay.io/coreos/etcd:v3.3.3
You could make the last form more permissive by telling it to only consider the imageOwner:
POST /api/image/quay.io/coreos/etcd:v3.3.3?filter=imageOwner
So that it would upgrade any resource using any etcd image regardless of whether or not it was the coreos Docker image or not.
Returns metadata for any resource that has an image matching the text supplied.
GET /api/resource/{image}
GET /api/resource/{repo}/{image}
GET /api/resource/{registry}/{repo}/{image}
The primary difference between this and the call for upgrade candidates is that this considers anything that matches whatever image segment is provided and returns a single list with no consideration given to upgrade eligibility.
It's just there to make it easy to:
When running as a service, all configuration is driven by environment variables:
K8S-URLK8S-HOSTK8S-TOKENK8S-CAK8S-CERTK8S-KEYK8S-USERNAMEK8S-PASSWORDFull argument names are shown in the command examples. Shorthand arguments are available, see the interactive CLI help to get a list.
npm i @arobson/hikaru -g
Only one of three forms is required. The user and password arguments (and optionally ca) will auth via basic. Using token (ca is also optional here) will auth via bearer token. Otherwise, the ca, cert and key are required to auth via certificates.
If tokens are present in a specification and not provided via a tokenFile, hikaru will prompt you for each of the tokens in turn. Skipping a token or proving a blank answer is not an option.
When deploying a spec that has already been deployed, hikaru will attempt to diff all manifests included and ignore identical resources while performing a rolling update (where available) on changed manifests.
saveDiffsA saveDiffs flag will cause any detected changes to be saved in a diff folder which includes the original, new and diff versions of each manifest that was updated in this way.
hikaru deploy {spec} \
--url {kubernetes url} \
--tokenFile {path to json, yaml or toml token file} \
--user {username} \
--password {password} \
--token {token} \
--ca {path to cluster CA} \
--cert {path to client cert} \
--key {path to client key} \
--saveDiffs \
--debug
hikaru deploy {spec} \
--url {kubernetes url} \
--tokenFile {path to json, yaml or toml token file} \
--user {username} \
--password {password} \
--token {token} \
--ca {path to cluster CA} \
--cert {path to client cert} \
--key {path to client key} \
--debug
The filter argument accepts a comma delimited list of fields used to determine upgrade eligibility. Valid fields are:
imageNameimageOwnerownerrepobranchfullVersionversionbuildcommithikaru upgrade {image} \
--url {kubernetes url} \
--filter {comma delimited filter option} \
--user {username} \
--password {password} \
--token {token} \
--ca {path to cluster CA} \
--cert {path to client cert} \
--key {path to client key} \
--debug
The filter argument accepts a comma delimited list of fields used to determine upgrade eligibility. Valid fields are:
imageNameimageOwnerownerrepobranchfullVersionversionbuildcommithikaru candidates {image} \
--url {kubernetes url} \
--filter {comma delimited filter option} \
--user {username} \
--password {password} \
--token {token} \
--ca {path to cluster CA} \
--cert {path to client cert} \
--key {path to client key} \
--debug
hikaru findByImage {image} \
--url {kubernetes url} \
--user {username} \
--password {password} \
--token {token} \
--ca {path to cluster CA} \
--cert {path to client cert} \
--key {path to client key} \
--debug
hikaru uses fount to supply dependencies on demand. This will affect how you use it to control configuration if you want to set it via a method other than the environment variables:
const hikaru = require('hikaru')
// make config changes *before* making any
// hikaru calls
const fount = require('fount')
const config = fount.get('config')
config.url = '' // the kubernetes url
config.username = '' // basic auth username
config.password = '' // basic auth password
config.token = '' // bearer token
config.ca = '' // kubernetes CA
config.cert = '' // kubernetes client cert
config.key = '' // kubernetes client key
// note: the ca, cert, and key would all
// need to be loaded from the file
// DO NOT provide a filename, it won't work
const hikaru = require('hikaru')
hikaru.deployCluster(
'git://github.com/me/my-spec',
{
branch: 'special', // 'master' is default
// kubernetes API/cluster version
version: '1.8', // '1.7' is default
data: {
token1: 'value'
token2: 100
}
})
.then(
() => {},
err => {}
)
const hikaru = require('hikaru')
hikaru.removeCluster(
'git://github.com/me/my-spec',
{
branch: 'special', // 'master' is default
// kubernetes API/cluster version
version: '1.8', // '1.7' is default
data: {
token1: 'value'
token2: 100
}
})
.then(
() => {},
err => {}
)
Note: the filter option defaults to:
['imageName', 'imageOwner', 'owner', 'repo', 'branch']
Which limits upgrades so that they'll only happen when the resource matches the image on these 5 fields.
This can be changed to be more or less permissive but should be done with great care. An empty filter should NEVER BE SUBMITTED as it would effectively be telling hikaru to replace all running pods with the same image.
const hikaru = require('hikaru')
hikaru.upgradeImage(
'myRepo/myImage:1.1.0',
{
filter: ['imageOwner', 'imageName']
})
.then(
list => {
// list of upgraded resources
},
err => {}
)
Works like the upgrade command but instead of performing the upgrade, only returns the list of resources that would be upgraded.
Very useful for checking to see which resources would be affected before running a command.
It is recommended to use this to check with users before running the actual upgrade - especially if they've been allowed to change any filter settings.
const hikaru = require('hikaru')
hikaru.getCandidatesImage(
'myRepo/myImage:1.1.0',
{
filter: ['imageOwner', 'imageName']
})
.then(
list => {
// list of upgraded resources
},
err => {}
)
Performs a search in the cluster for resources with an image containing the provided string.
Useful for trying to see which things might be in use across various owners. Example: searching for nginx to see how many different NGiNX containers might be deployed.
const hikaru = require('hikaru')
hikaru.findResources('nginx')
.then(
list => {
// list of matching resources
},
err => {}
)
Caveat Emptor: I do not have time (presently) to document the extensive cluster and k8s module APIs or provide adequate test coverage for them. You are using them at your own risk.
const hikaru = require('hikaru')
hikaru.connect()
.then(() => {
// both `.k8s` and `.cluster` will be available
// on hikaru after the connect promise resolves
})
A Docker image containing the hikaru HTTP service is already built for ease of use:
npmwharf/hikaru:latest
And also released with various version tagging schemes.
To use it, you can set the various environment variables to control behavior. To use it inside a Kubernetes cluster, I recommend pulling in the token and CA token that Kubernetes places in the pods. See the Kubernetes spec (below) for reference.
A mcgonagall kubernetes spec to deploy hikaru to a kubernetes cluster can be found at https://github.com/npm-wharf/hikaru-spec.
It's likely you'll want to copy hikaru.toml into your own full cluster specification to get the benefit of mgconagall's NGiNX generation. If that's not something you want though, you can simply deploy directly from the github repo using the CLI:
hikaru deploy git://github.com/npm-wharf/hikaru-spec \
-k {your cluster endpoint} \
-u {username} \
-p {password}
Note: there are other auth methods available, the cert approach is probably best :)
FAQs
A continuous deployment service and integration helper for kubernetes
The npm package @arobson/hikaru receives a total of 1 weekly downloads. As such, @arobson/hikaru popularity was classified as not popular.
We found that @arobson/hikaru demonstrated a not healthy version release cadence and project activity because the last version was released a year ago. It has 1 open source maintainer collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Research
Security News
Socket researchers uncover a malicious npm package posing as a tool for detecting vulnerabilities in Etherium smart contracts.
Security News
Research
A supply chain attack on Rspack's npm packages injected cryptomining malware, potentially impacting thousands of developers.
Research
Security News
Socket researchers discovered a malware campaign on npm delivering the Skuld infostealer via typosquatted packages, exposing sensitive data.