ChartMuseum is an open-source Helm Chart Repository server written in Go (Golang), with support for cloud storage backends, including Google Cloud Storage, Amazon S3, Microsoft Azure Blob Storage, Alibaba Cloud OSS Storage, Openstack Object Storage, Oracle Cloud Infrastructure Object Storage, Baidu Cloud BOS Storage, Tencent Cloud Object Storage, DigitalOcean Spaces, Minio, and etcd.

Works as a valid Helm Chart Repository, and also provides an API for uploading charts.

Powered by some great Go technology:

• helm/helm - for working with charts
• gin-gonic/gin - for HTTP routing
• urfave/cli - for command line option parsing
• spf13/viper - for configuration
• uber-go/zap - for logging
• chartmuseum/auth - for auth
• chartmuseum/storage - for multi-cloud storage

• GET /index.yaml - retrieved when you run helm repo add chartmuseum http://localhost:8080/
• GET /charts/mychart-0.1.0.tgz - retrieved when you run helm install chartmuseum/mychart
• GET /charts/mychart-0.1.0.tgz.prov - retrieved when you run helm install with the --verify flag

• POST /api/charts - upload a new chart version
• POST /api/prov - upload a new provenance file
• DELETE /api/charts/<name>/<version> - delete a chart version (and corresponding provenance file)
• GET /api/charts - list all charts
• GET /api/charts/<name> - list all versions of a chart
• GET /api/charts/<name>/<version> - describe a chart version
• GET /api/charts/<name>/<version>/templates - get chart template
• GET /api/charts/<name>/<version>/values - get chart values
• HEAD /api/charts/<name> - check if chart exists (any versions)
• HEAD /api/charts/<name>/<version> - check if chart version exists

• GET / - HTML welcome page
• GET /info - returns current ChartMuseum version
• GET /health - returns 200 OK

_{Follow "How to Run" section below to get ChartMuseum up and running at http://localhost:8080}

First create mychart-0.1.0.tgz using the Helm CLI:

cd mychart/
helm package .

Upload mychart-0.1.0.tgz:

curl --data-binary "@mychart-0.1.0.tgz" http://localhost:8080/api/charts

If you've signed your package and generated a provenance file, upload it with:

curl --data-binary "@mychart-0.1.0.tgz.prov" http://localhost:8080/api/prov

Both files can also be uploaded at once (or one at a time) on the /api/charts route using the multipart/form-data format:

curl -F "[email protected]" -F "[email protected]" http://localhost:8080/api/charts

You can also use the helm-push plugin:

helm cm-push mychart/ chartmuseum

Add the URL to your ChartMuseum installation to the local repository list:

helm repo add chartmuseum http://localhost:8080

Search for charts:

helm search repo chartmuseum/

Install chart:

helm install chartmuseum/mychart --generate-name

You can use the installer script:

curl https://raw.githubusercontent.com/helm/chartmuseum/main/scripts/get-chartmuseum | bash

or download manually from the releases page, which also contains all package checksums and signatures.

Determine your version with chartmuseum --version.

Show all CLI options with chartmuseum --help. Common configurations can be seen below.

All command-line options can be specified as environment variables, which are defined by the command-line option, capitalized, with all -'s replaced with _'s.

For example, the env var STORAGE_AMAZON_BUCKET can be used in place of --storage-amazon-bucket.

Use chartmuseum --config config.yaml to read configuration from a file.

When using file-based configuration, the corresponding option name can be looked up in pkg/config/vars.go. It would be the key of configVars entry corresponding to the command line option / environment variable. For example, --storage corresponds to storage.backend in the configuration file.

Here's a complete example of a config.yaml:

debug: true
port: 8080
storage.backend: local
storage.local.rootdir: <storage_path>
bearerauth: 1
authrealm: <authorization server url>
authservice: <authorization server service name>
authcertpath: <path to authorization server public pem file>
authactionssearchpath: <optional: JMESPath to find allowed actions in a jwt token>
depth: 2

Make sure your environment is properly setup to access my-s3-bucket

For Amazon S3, endpoint is automatically inferred.

chartmuseum --debug --port=8080 \
  --storage="amazon" \
  --storage-amazon-bucket="my-s3-bucket" \
  --storage-amazon-prefix="" \
  --storage-amazon-region="us-east-1"

For S3 compatible services like Minio, set the credentials using environment variables and pass the endpoint.

export AWS_ACCESS_KEY_ID=""
export AWS_SECRET_ACCESS_KEY=""
chartmuseum --debug --port=8080 \
  --storage="amazon" \
  --storage-amazon-bucket="my-s3-bucket" \
  --storage-amazon-prefix="" \
  --storage-amazon-region="us-east-1" \
  --storage-amazon-endpoint="my-s3-compatible-service-endpoint"

You need at least the following permissions inside your IAM Policy

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "AllowListObjects",
      "Effect": "Allow",
      "Action": [
        "s3:ListBucket"
      ],
      "Resource": "arn:aws:s3:::my-s3-bucket"
    },
    {
      "Sid": "AllowObjectsCRUD",
      "Effect": "Allow",
      "Action": [
        "s3:DeleteObject",
        "s3:GetObject",
        "s3:PutObject"
      ],
      "Resource": "arn:aws:s3:::my-s3-bucket/*"
    }
  ]
}

In order to work with AWS service accounts you may need to set AWS_SDK_LOAD_CONFIG=1 in your environment. For more context, please see here.

If you are using S3-Compatible storage, provider of S3 storage has disabled path-style and force virtual hosted-style, you can use specify storage-amazon-force-path-style options as following example:

export AWS_ACCESS_KEY_ID=""
export AWS_SECRET_ACCESS_KEY=""
chartmuseum --debug --port=8080 \
  --storage="amazon" \
  --storage-amazon-bucket="my-s3-bucket" \
  --storage-amazon-prefix="" \
  --storage-amazon-region="us-east-1" \
  --storage-amazon-endpoint="my-s3-compatible-service-endpoint"
  --storage-amazon-force-path-style=false

For DigitalOcean, set the credentials using environment variable and pass the endpoint. Note below, that the region us-east-1 needs to be set, since that is how the DigitalOcean cli implementation functions. The actual region of your spaces location is defined by the endpoint. Below we are using Frankfurt as an example.

export AWS_ACCESS_KEY_ID="spaces_access_key"
export AWS_SECRET_ACCESS_KEY="spaces_secret_key"
  chartmuseum --debug --port=8080 \
  --storage="amazon" \
  --storage-amazon-bucket="my_spaces_name" \
  --storage-amazon-prefix="my_spaces_name_subfolder" \
  --storage-amazon-region="us-east-1" \
  --storage-amazon-endpoint="https://fra1.digitaloceanspaces.com"

The access_key and secret_key can be generated from the DigitalOcean console, under the section API/Spaces_access_keys.

Note: on certain S3-based storage backends, the LastModified field on objects is truncated to the nearest second. For more info, please see issue #152.

In order to mitigate this, you may use use the --storage-timestamp-tolerance option. For example, to round to the nearest second, you could use --storage-timestamp-tolerance=1s. For acceptable values to use for this field, please see here.

Make sure your environment is properly setup to access my-gcs-bucket.

One way to do so is to set the GOOGLE_APPLICATION_CREDENTIALS var in your environment, pointing to the JSON file containing your service account key:

export GOOGLE_APPLICATION_CREDENTIALS="/home/user/Downloads/[FILE_NAME].json"

More info on Google Cloud authentication can be found here.

chartmuseum --debug --port=8080 \
  --storage="google" \
  --storage-google-bucket="my-gcs-bucket" \
  --storage-google-prefix=""

Make sure your environment is properly setup to access mycontainer.

To do so, you must set the following env vars:

• AZURE_STORAGE_ACCOUNT
• AZURE_STORAGE_ACCESS_KEY

chartmuseum --debug --port=8080 \
  --storage="microsoft" \
  --storage-microsoft-container="mycontainer" \
  --storage-microsoft-prefix=""

Make sure your environment is properly setup to access my-oss-bucket.

To do so, you must set the following env vars:

• ALIBABA_CLOUD_ACCESS_KEY_ID
• ALIBABA_CLOUD_ACCESS_KEY_SECRET

chartmuseum --debug --port=8080 \
  --storage="alibaba" \
  --storage-alibaba-bucket="my-oss-bucket" \
  --storage-alibaba-prefix="" \
  --storage-alibaba-endpoint="oss-cn-beijing.aliyuncs.com"

Make sure your environment is properly setup to access mycontainer.

To do so, you must set the following env vars (depending on your openstack version):

• OS_AUTH_URL
• either OS_PROJECT_NAME or OS_TENANT_NAME or OS_PROJECT_ID or OS_TENANT_ID
• either OS_DOMAIN_NAME or OS_DOMAIN_ID
• either OS_USERNAME or OS_USERID
• OS_PASSWORD

chartmuseum --debug --port=8080 \
  --storage="openstack" \
  --storage-openstack-container="mycontainer" \
  --storage-openstack-prefix="" \
  --storage-openstack-region="myregion"

For Swift V1 Auth you must set the following env vars:

• ST_AUTH
• ST_USER
• ST_KEY

chartmuseum --debug --port=8080 \
  --storage="openstack" \
  --storage-openstack-auth="v1" \
  --storage-openstack-container="mycontainer" \
  --storage-openstack-prefix=""

Make sure your environment is properly setup to access my-ocs-bucket.

More info on Oracle Cloud Infrastructure authentication can be found here.

chartmuseum --debug --port=8080 \
  --storage="oracle" \
  --storage-oracle-bucket="my-ocs-bucket" \
  --storage-oracle-prefix="" \
  --storage-oracle-compartmentid="ocid1.compartment.oc1..1234"

Make sure your environment is properly setup to access my-bos-bucket.

To do so, you must set the following env vars:

• BAIDU_CLOUD_ACCESS_KEY_ID
• BAIDU_CLOUD_ACCESS_KEY_SECRET

chartmuseum --debug --port=8080 \
  --storage="baidu" \
  --storage-baidu-bucket="my-bos-bucket" \
  --storage-baidu-prefix="" \
  --storage-baidu-endpoint="bj.bcebos.com"

Make sure your environment is properly setup to access my-cos-bucket.

To do so, you must set the following env vars:

• TENCENT_CLOUD_COS_SECRET_ID
• TENCENT_CLOUD_COS_SECRET_KEY

chartmuseum --debug --port=8080 \
  --storage="tencent" \
  --storage-tencent-bucket="my-cos-bucket" \
  --storage-tencent-prefix="" \
  --storage-tencent-endpoint="cos.ap-beijing.myqcloud.com"

To use etcd as backend you need the CA certificate and the signed key pair. See here

chartmuseum --debug --port=8080 \
  --storage="etcd" \
  --storage-etcd-cafile="/path/to/ca.crt" \
  --storage-etcd-certfile="/path/to/server.crt" \
  --storage-etcd-keyfile="/path/to/server.key" \
  --storage-etcd-prefix="" \
  --storage-etcd-endpoint="http://localhost:2379"

Make sure you have read-write access to ./chartstorage (will create if doesn't exist on first upload)

chartmuseum --debug --port=8080 \
  --storage="local" \
  --storage-local-rootdir="./chartstorage"

If both of the following options are provided, basic http authentication will protect all routes:

• --basic-auth-user=<user> - username for basic http authentication
• --basic-auth-pass=<pass> - password for basic http authentication

You may want basic auth to only be applied to operations that can change Charts, i.e. PUT, POST and DELETE. So to avoid basic auth on GET operations use

• --auth-anonymous-get - allow anonymous GET operations

If all of the following options are provided, bearer auth will protect all routes:

• --bearer-auth - enables bearer auth
• --auth-realm=<realm> - authorization server url
• --auth-service=<service> - authorization server service name
• --auth-cert-path=<path> - path to authorization server public pem file
• --auth-actions-search-path=<JMESPath> - (optional) JMESPath to find allowed actions in a jwt token

Using options above, ChartMuseum is configured with a public key, and will accept RS256 JWT tokens signed by the associated private key, passed in the Authorization header. You can use the chartmuseum/auth Go library to generate valid JWT tokens.

In order to gain access to a specific resource, the JWT token must contain an access section in the claims. This section indicates which resources the user is able to access. Here is an example token payload:

{
  "exp": 1543995770,
  "iat": 1543995470,
  "access": [
    {
      "type": "artifact-repository",
      "name": "org1/repo1",
      "actions": [
        "pull"
      ]
    }
  ]
}

The type is always "artifact-repository", the name is the namespace/tenant (just use the string "repo" if using single-tenant server), and actions is an array of actions the user can perform ("pull" and/or "push).

If your JWT token structure is different, you can configure a JMESPath string. So you can define the way to find the allowed actions yourself. For the type and the the name you can use following placeholder

• name: $NAMESPACE
• type: $ACCESS_ENTRY_TYPE

E.g.: If you want to represent the default configuration, the JMESPath looks like: access[?name=='$NAMESPACE' && type=='$ACCESS_ENTRY_TYPE'].actions[].

For more information about how this works, please see chartmuseum/auth-server-example.

If both of the following options are provided, the server will listen and serve HTTPS:

• --tls-cert=<crt> - path to tls certificate chain file
• --tls-key=<key> - path to tls key file

If the above HTTPS values are provided in addition to below, the server will listen and serve HTTPS and authenticate client requests against the CA certificate:

• --tls-ca-cert=<cacert> - path to tls certificate file

You can specify the --gen-index option if you only wish to use ChartMuseum to generate your index.yaml file. Note that this will only work with --depth=0.

The contents of index.yaml will be printed to stdout and the program will exit. This is useful if you are satisfied with your current Helm CI/CD process and/or don't want to monitor another webservice.

• --log-json - output structured logs as json
• --log-health - log incoming /health requests
• --log-latency-integer - log latency as an integer (nanoseconds) instead of a string
• --disable-api - disable all routes prefixed with /api
• --disable-delete - explicitly disable the delete chart route
• --disable-statefiles - disable use of index-cache.yaml
• --allow-overwrite - allow chart versions to be re-uploaded without ?force querystring
• --disable-force-overwrite - do not allow chart versions to be re-uploaded, even with ?force querystring
• --chart-url=<url> - absolute url for .tgzs in index.yaml
• --storage-amazon-endpoint=<endpoint> - alternative s3 endpoint
• --storage-amazon-sse=<algorithm> - s3 server side encryption algorithm
• --storage-openstack-cacert=<path> - path to a custom ca certificates bundle for openstack
• --chart-post-form-field-name=<field> - form field which will be queried for the chart file content
• --prov-post-form-field-name=<field> - form field which will be queried for the provenance file content
• --index-limit=<number> - limit the number of parallel indexers
• --context-path=<path> - base context path (new root for application routes)
• --depth=<number> - levels of nested repos for multitenancy
• --cors-alloworigin=<value> - value to set in the Access-Control-Allow-Origin HTTP header
• --read-timeout=<number> - socket read timeout for http server
• --write-timeout=<number> - socker write timeout for http server

Available via GitHub Container Registry (GHCR).

Example usage (local storage):

docker run --rm -it \
  -p 8080:8080 \
  -e DEBUG=1 \
  -e STORAGE=local \
  -e STORAGE_LOCAL_ROOTDIR=/charts \
  -v $(pwd)/charts:/charts \
  ghcr.io/helm/chartmuseum:v0.16.2

Example usage (S3):

docker run --rm -it \
  -p 8080:8080 \
  -e DEBUG=1 \
  -e STORAGE="amazon" \
  -e STORAGE_AMAZON_BUCKET="my-s3-bucket" \
  -e STORAGE_AMAZON_PREFIX="" \
  -e STORAGE_AMAZON_REGION="us-east-1" \
  -v ~/.aws:/home/chartmuseum/.aws:ro \
  ghcr.io/helm/chartmuseum:v0.16.2

There is a Helm chart for ChartMuseum itself.

You can also view it on Artifact Hub.

To install:

helm repo add chartmuseum https://chartmuseum.github.io/charts
helm install chartmuseum/chartmuseum

If interested in making changes, please submit a PR to chartmuseum/charts. Before doing any work, please check for any currently open pull requests. Thanks!

Multitenancy is supported with the --depth flag.

To begin, start with a directory structure such as

charts
├── org1
│   ├── repoa
│   │   └── nginx-ingress-0.9.3.tgz
├── org2
│   ├── repob
│   │   └── chartmuseum-0.4.0.tgz

This represents a storage layout appropriate for --depth=2. The organization level can be eliminated by using --depth=1. The default depth is 0 (singletenant server).

Start the server with --depth=2, pointing to the charts/ directory:

chartmuseum --debug --depth=2 --storage="local" --storage-local-rootdir=./charts

This example will provide two separate Helm Chart Repositories at the following locations:

• http://localhost:8080/org1/repoa
• http://localhost:8080/org2/repob

This should work with all supported storage backends.

To use the chart manipulation routes, simply place the name of the repo directly after "/api" in the route:

curl -F "[email protected]" http://localhost:8080/api/org1/repoa/charts

You may also experiment with the --depth-dynamic flag, which should allow for dynamic depth levels (i.e. all of /api/charts, /api/myrepo/charts, /api/org1/repoa/charts).

For large chart repositories, you may wish to paginate the results from the GET /api/charts route.

To do so, add the offset and limit query params to the request. For example, to retrieve a list of 5 charts total, skipping the first 5 charts, you could use the following:

GET /api/charts?offset=5&limit=5

By default, the contents of index.yaml (per-tenant) will be stored in memory. This means that memory usage will continue to grow indefinitely as more charts are added to storage.

You may wish to offload this to an external cache store, especially for large, multitenant installations.

When dealing with thousands of charts, you may experience latency with the default settings. This is because upon each request, the storage backend is scanned for changes compared to the cache.

If you are ok with index.yaml being out-of-date for a fixed period of time, you can improve performance by using the --cache-interval=<interval> option. When this setting is enabled, the charts available for each tenant are refreshed on a timer.

For example, to only check storage every 5 minutes, you can use --cache-interval=5m.

For valid values to use for this setting, please see here.

Example of using Redis as an external cache store:

chartmuseum --debug --port=8080 \
  --storage="local" \
  --storage-local-rootdir="./chartstorage" \
  --cache="redis" \
  --cache-redis-addr="localhost:6379" \
  --cache-redis-password="" \
  --cache-redis-db=0

ChartMuseum exposes its Prometheus metrics at the /metrics route on the main port. This can be enabled with the --enable-metrics command-line flag or the ENABLE_METRICS environment variable.

Note that the Kubernetes chart currently disables metrics by default (ENABLE_METRICS=false is set in the chart). The --disable-metrics command-line flag has be deprecated and will only be available in v0.14.0 and prior.

Below are the current application metrics exposed. Note that there is a per tenant (repo) label. The repo label corresponds to the depth parameter, so a depth=2 as the example above would have repo labels named org1/repoa and org2/repob.

*: see above for repo label

There are other general global metrics harvested (per process, hence for all tenants). You can get the complete list by using the /metrics route.

The repository index (index.yaml) is dynamically generated based on packages found in storage. If you store your own version of index.yaml, it will be completely ignored.

GET /index.yaml occurs when you run helm repo add chartmuseum http://localhost:8080 or helm repo update.

If you manually add/remove a .tgz package from storage, it will be immediately reflected in GET /index.yaml.

You are no longer required to maintain your own version of index.yaml using helm repo index --merge.

The --gen-index CLI option (described above) can be used to generate and print index.yaml to stdout.

Upon index regeneration, ChartMuseum will, however, save a statefile in storage called index-cache.yaml used for cache optimization. This file is only meant for internal use, but may be able to be used for migration to simple storage.

Please see scripts/mirror-k8s-repos.sh for an example of how to download all .tgz packages from the official Kubernetes repositories (both stable and incubator).

You can then use ChartMuseum to serve up an internal mirror:

scripts/mirror-k8s-repos.sh
chartmuseum --debug --port=8080 --storage="local" --storage-local-rootdir="./mirror"

With the flag --web-template-path=<path>, you can specify the path to your custom welcome page.

The structure of the folder should be like this:

web/
  index.html
  xyz.html
  static/
      main.css
      main.js

ChartMuseum is using gin-gonic to serve the static files, this means that you can use go-template to render the files.

If you don't specify a custom welcome page, ChartMuseum will serve the default one.

By setting the flag --artifact-hub-repo-id <repo id>, ChartMuseum will serve a artifacthub-repo.yml file with the specified repo ID in the repositoryID field of the yaml file.

repositoryID: The ID of the Artifact Hub repository where the packages will be published to (optional, but it enables verified publisher)

For multitenancy setups, you can provide a key value pair to the flag in the format: --artifact-hub-repo-id <repo>=<repo id>

chartmuseum --storage local --storage-local-rootdir /tmp/ --depth 1 --artifact-hub-repo-id org1=<repo id> --artifact-hub-repo-id org2=<repo2 id>

The artifacthub-repo.yml file will then be served at /org1/artifacthub-repo.yml and /org2/artifacthub-repo.yml

_{"Preserve your precious artifacts... in the cloud!"}

The following subprojects are maintained by ChartMuseum:

• chartmuseum/helm-push - Helm plugin to push chart package to ChartMuseum
• chartmuseum/storage - Go library providing common interface for working across multiple cloud storage backends
• chartmuseum/auth - Go library for generating ChartMuseum JWT Tokens, authorizing HTTP requests, etc.
• chartmuseum/auth-server-example - Example server providing JWT tokens for ChartMuseum auth
• chartmuseum/testbed - Docker testbed for continuous integration
• chartmuseum/www - chartmuseum.com static site source code
• chartmuseum/ui - ChartMuseum frontend UI

You can reach the ChartMuseum community and developers in the Kubernetes Slack #chartmuseum channel.