anchore-k8s-inventory
polls the Kubernetes API on an interval to retrieve which images are currently in use.
It can be run inside a cluster (under a Service Account) or outside (via any provided kubeconfig).
Install the binary or Download the Docker image
anchore-k8s-inventory
can be run as a CLI, Docker Container, or Helm Chart
By default, anchore-k8s-inventory
will look for a kubeconfig in the home directory to use to authenticate (when run as a CLI).
$ anchore-k8s-inventory --verbose-inventory-reports
{
"cluster_name": "docker-desktop",
"containers": [
{
"id": "docker://911d2cf6351cbafc349f131aeef1b1fb295a889504d38c89a065da1a91d828b9",
"image_digest": "sha256:76049887f07a0476dc93efc2d3569b9529bf982b22d29f356092ce206e98765c",
"image_tag": "docker.io/kubernetesui/metrics-scraper:v1.0.8",
"name": "dashboard-metrics-scraper",
"pod_uid": "c5b40099-20a5-4b46-8062-cf84f9d6ac23"
},
{
"id": "docker://a9cd75ad99dd4363bbd882b40e753b58c62bfd7b03cabeb764c1dac97568ad26",
"image_digest": "sha256:2e500d29e9d5f4a086b908eb8dfe7ecac57d2ab09d65b24f588b1d449841ef93",
"image_tag": "docker.io/kubernetesui/dashboard:v2.7.0",
"name": "kubernetes-dashboard",
"pod_uid": "72ba7e4e-6e35-48c0-bff7-558a525074d5"
},
.....
],
"namespaces": [
{
"labels": {
"kubernetes.io/metadata.name": "kube-public"
},
"name": "kube-public",
"uid": "dd561bf1-11ff-4381-8a1f-f156c206fe13"
},
{
"labels": {
"kubernetes.io/metadata.name": "kube-system"
},
"name": "kube-system",
"uid": "012ebe67-dd49-4fd9-b604-258385df3957"
},
.....
],
"nodes": [
{
"annotations": {
"kubeadm.alpha.kubernetes.io/cri-socket": "unix:///var/run/cri-dockerd.sock",
"node.alpha.kubernetes.io/ttl": "0",
"volumes.kubernetes.io/controller-managed-attach-detach": "true"
},
"arch": "arm64",
"container_runtime_version": "docker://20.10.23",
"kernel_version": "5.15.49-linuxkit",
"kube_proxy_version": "v1.26.1",
"kubelet_version": "v1.26.1",
"labels": {
"beta.kubernetes.io/arch": "arm64",
"beta.kubernetes.io/os": "linux",
"kubernetes.io/arch": "arm64",
"kubernetes.io/hostname": "minikube",
"kubernetes.io/os": "linux",
"minikube.k8s.io/commit": "ddac20b4b34a9c8c857fc602203b6ba2679794d3",
"minikube.k8s.io/name": "minikube",
"minikube.k8s.io/primary": "true",
"minikube.k8s.io/updated_at": "2023_04_11T11_20_54_0700",
"minikube.k8s.io/version": "v1.29.0",
"node-role.kubernetes.io/control-plane": "",
"node.kubernetes.io/exclude-from-external-load-balancers": ""
},
"name": "minikube",
"operating_system": "linux",
"uid": "b8334e25-68a5-4cbc-bf7a-fc188f2c6023"
}
],
"pods": [
{
"annotations": {
"seccomp.security.alpha.kubernetes.io/pod": "runtime/default"
},
"labels": {
"k8s-app": "dashboard-metrics-scraper",
"pod-template-hash": "5c6664855"
},
"name": "dashboard-metrics-scraper-5c6664855-s8lpc",
"namespace_uid": "c1d98ff5-6689-4016-aef3-8802790c3b10",
"node_uid": "b8334e25-68a5-4cbc-bf7a-fc188f2c6023",
"uid": "c5b40099-20a5-4b46-8062-cf84f9d6ac23"
},
{
"labels": {
"gcp-auth-skip-secret": "true",
"k8s-app": "kubernetes-dashboard",
"pod-template-hash": "55c4cbbc7c"
},
"name": "kubernetes-dashboard-55c4cbbc7c-6p28m",
"namespace_uid": "c1d98ff5-6689-4016-aef3-8802790c3b10",
"node_uid": "b8334e25-68a5-4cbc-bf7a-fc188f2c6023",
"uid": "72ba7e4e-6e35-48c0-bff7-558a525074d5"
},
.....
],
"serverVersionMetadata": {
"major": "1",
"minor": "26",
"gitVersion": "v1.26.1",
"gitCommit": "8f94681cd294aa8cfd3407b8191f6c70214973a4",
"gitTreeState": "clean",
"buildDate": "2023-01-18T15:51:25Z",
"goVersion": "go1.19.5",
"compiler": "gc",
"platform": "linux/arm64"
},
"timestamp": "2023-05-03T12:34:13Z"
}
In order to run anchore-k8s-inventory
as a container, it needs a kubeconfig
~ docker run -it --rm -v ~/.kube/config:/.kube/config anchore/k8s-inventory:latest --verbose-inventory-reports
Anchore-k8s-inventory is the foundation of Anchore Enterprise's Runtime Inventory feature. Running anchore-k8s-inventory via Helm is a great way to retrieve your Kubernetes Image inventory without providing Cluster Credentials to Anchore.
Anchore-k8s-inventory runs as a read-only service account in the cluster it's deployed to.
In order to report the inventory to Anchore, anchore-k8s-inventory does require authentication material for your Anchore Enterprise deployment. anchore-k8s-inventory's helm chart automatically creates a kubernetes secret for the Anchore Password based on the values file you use, Ex.:
anchore-k8s-inventory:
anchore:
password: foobar
It will set the following environment variable based on this: ANCHORE_K8S_INVENTORY_ANCHORE_PASSWORD=foobar
.
If you don't want to store your Anchore password in the values file, you can create your own secret to do this:
apiVersion: v1
kind: Secret
metadata:
name: anchore-k8s-inventory-anchore-password
type: Opaque
stringData:
ANCHORE_K8S_INVENTORY_ANCHORE_PASSWORD: foobar
and then provide it to the helm chart via the values file:
anchore-k8s-inventory:
existingSecret: anchore-k8s-inventory-anchore-password
anchore-k8s-inventory's helm chart is part of the charts.anchore.io repo. You can install it via:
helm repo add anchore https://charts.anchore.io
helm install <release-name> -f <values.yaml> anchore/k8s-inventory
A basic values file can always be found here
In the Enterprise v5.11.0
release, Enterprise provides the ability for k8s-inventory to register and provide health
reporting. This allows Enterprise to track the status of the agents and expose that information to administrators.
In the API that Enterprise provides for this feature, the entities are referred to as Integrations. Hence, the
anchore-k8s-inventory agent is a type of Integration.
Before an agent can start sending health reports, it needs to register with Enterprise. This is a handshake whereby the agent presents itself to Enterprise. The registration contains information such as type of integration, its name, when it was started, uptime, how often it will send health reports etc. Some of the properties are optional such as the 'description' property.
When the agent registers it is assigned an integration uuid by Enterprise. This uuid is then used for all API operations that target that particular integration instance. Hence, when the agent sends its health reports to Enterprise, it makes a REST API call to a URL that includes the integration uuid.
When the agent registers with Enterprise, the registration will include a registration_id
and a
registration_instance_id
. These are used by Enterprise to look up the integration uuid for the agent.
If it is the first time the agent registers, no integration uuid
will exist so Enterprise will create one and store the
association between the integration uuid and <registration_id, registration_instance_id>
in its database. For
subsequent re-registrations (e.g., if the agent pod is restarted), the association will allow Enterprise to return the
integration uuid
created in the earlier registration. Hence, as long as the
<registration_id, registration_instance_id>
pair remains the same, Enterprise will consider it to be the same
integration instance. The reason a value pair is needed is that integrations like the anchore_k8s_inventory
agent can
be deployed as multiple replicas (in the case of K8s Deployment). Enterprise must be able to differentiate between the
agent replica instances. A value pair with sensibly chosen values makes that possible.
The registration_id
can be set via configuration (see the Configuration section below).
Only the agent itself can set the registration_instance_id
value. It will set it to the hostname where the agent runs
(or if its empty, generate a uuid and use that value).
If the agent interacts with an Enterprise deployment that does not support Integration registration and health
reporting (i.e., Enterprise releases < v5.11.0
), it will skip registration, disable health reporting and then let
inventory reporting continue like in pre-v1.7.0
releases. The agent will periodically check if Enterprise has been
upgraded to v5.11.0
or later and perform registration and enable health reporting in such cases. No restart of the
agent is required for this to happen.
# same as -q ; suppress all output (except for the inventory results)
quiet: false
log:
# use structured logging
structured: false
# the log level; note: detailed logging suppress the ETUI
level: "debug"
# location to write the log file (default is not to have a log file)
file: "./anchore-k8s-inventory.log"
# enable/disable checking for application updates on startup
check-for-app-update: true
kubeconfig:
path:
cluster: docker-desktop
cluster-cert:
server: # ex. https://kubernetes.docker.internal:6443
user:
type: # valid: [private_key, token]
client-cert:
private-key:
token:
# enable/disable printing inventory reports to stdout
verbose-inventory-reports: false
Configure values for the registration of the agent as an Integration.
The registration_id
can preferably be left empty if the Anchore helm chartsk8s-inventory v0.5.0
or later are used.
If explicitly set in the configuration it is advisable to use some tool that can generate uuids to ensure uniqueness.
anchore-registration:
# The id to register the agent as with Enterprise, so Enterprise can map the agent to its integration uuid.
# If left unspecified, the agent will attempt to set registration-id to the uid of the K8s Deployment for the agent.
# If that fails (e.g., if the agent is not deployed on K8s), the agent will generate a UUID to use as registration-id.
registration-id:
# The name that the agent should have. If left unspecified, the agent will attempt to set it to the name of the K8s
# Deployment for the agent. If that fails it will be empty.
integration-name:
# A short description for the agent
integration-description:
Configure which namespaces anchore-k8s-inventory should search.
include
section- A list of explicit strings that will detail the list of namespaces to capture image data from.
- If left as an empty list
[]
all namespaces will be searched - Example:
namespace-selectors:
include:
- default
- kube-system
- prod-app
exclude
section- A list of explicit strings and/or regex patterns for namespaces to be excluded.
- A regex is determined if the string does not match standard DNS name requirements.
- Example:
namespace-selectors:
exclude:
- default
- ^kube-*
- ^prod-*
# Which namespaces to search or exclude.
namespace-selectors:
# Namespaces to include as explicit strings, not regex
# NOTE: Will search ALL namespaces if left as an empty array
include: []
# List of namespaces to exclude, can use explicit strings and/or regexes.
# For example
#
# list:
# - default
# - ^kube-*
#
# Will exclude the default, kube-system, and kube-public namespaces
exclude: []
# If true then namespaces containing 0 pods will be omitted from the report sent to Anchore Enterprise
ignore-empty: false
The following configuration options can determine which Anchore account
inventory reports are sent to. Without any of the following configuration the
account set in the anchore
section will be used.
If a mixture of static account routing and account routing by namespace label is used then the static account routes configured in k8s-inventory config will take precedence over any account that is specified by namespace label.
Set a list of accounts and which namespaces inventory should be sent to that
account. You can override the default credentials on a per account basis, if
not set then the global credentials set in the anchore
section will be used.
account-routes:
# <Anchore Account Name>: # (this is the name of the anchore account e.g. admin)
# user: <username> <OPTIONAL>
# password: <password> <OPTIONAL>
# namespaces: # Can be a list of explicit namespaces matches or regex patterns
# - <namespace>
# - <regex pattern>
#
# Example
# admin:
# user: username
# password: password
# namespaces:
# - default
# - ^kube-*
In this mode use a label set on a kubernetes namespace to determine which
Anchore account inventory data for that namespace should be sent to. It is
assumed that the credentials set in the anchore
section can post to all
accounts.
# Route namespaces to anchore accounts by a label on the namespace
account-route-by-namespace-label:
# The name of the namespace label that will be used to route the contents of
# that namespace to the Anchore account matching the value of the label
key: # e.g anchore.io/account.name
# The name of the account to route inventory to for a namespace that is
# missing the label or if the anchore account is not found.
# If not set then it will default to the account specified in the anchore credentials
default-account: # e.g. admin
# If true will exclude inventorying namespaces that are missing the specified label
ignore-namespace-missing-label: false
This section will allow users to tune the way anchore-k8s-inventory interacts with the kubernetes API server.
# Kubernetes API configuration parameters (should not need tuning)
kubernetes:
# Sets the request timeout for kubernetes API requests
request-timeout-seconds: 60
# Sets the number of objects to iteratively return when listing resources
request-batch-size: 100
# Worker pool size for collecting pods from namespaces. Adjust this if the api-server gets overwhelmed
worker-pool-size: 100
# Can be one of adhoc, periodic (defaults to adhoc)
mode: adhoc
# Only respected if mode is periodic
polling-interval-seconds: 300
There are cases where images in Kubernetes do not have an associated tag - for example when an image is deployed using the digest.
kubectl run python --image=python@sha256:f0a210a37565286ecaaac0529a6749917e8ea58d3dfc72c84acfbfbe1a64a20a
Anchore Enterprise will use the image digest to process an image but it still requires a tag to be
associated with the image. The missing-tag-policy
lets you configure the best way to handle the
missing tag edge case in your environment.
digest will use the image digest as a dummy tag.
{
"tag": "alpine:4ed1812024ed78962a34727137627e8854a3b414d19e2c35a1dc727a47e16fba",
"repoDigest": "sha256:4ed1812024ed78962a34727137627e8854a3b414d19e2c35a1dc727a47e16fba"
}
insert will use a dummy tag configured by missing-tag-policy.tag
{
"tag": "alpine:UNKNOWN",
"repoDigest": "sha256:4ed1812024ed78962a34727137627e8854a3b414d19e2c35a1dc727a47e16fba"
}
drop will simply ignore the images that don't have tags.
# Handle cases where a tag is missing. For example - images designated by digest
missing-tag-policy:
# One of the following options [digest, insert, drop]. Default is 'digest'
#
# [digest] will use the image's digest as a dummy tag.
#
# [insert] will insert a default tag in as a dummy tag. The dummy tag is
# customizable under missing-tag-policy.tag
#
# [drop] will drop images that do not have tags associated with them. Not
# recommended.
policy: digest
# Dummy tag to use. Only applicable if policy is 'insert'. Defaults to UNKNOWN
tag: UNKNOWN
# Ignore images out of pods that are not in a Running state
ignore-not-running: true
Set upper limits for the content that can be contained in a single inventory report POST to Anchore Enterprise. If the inventory data is greater than the limit then the inventory report will be broken into smaller requests up to the limit size specified.
inventory-report-limits:
namespaces: 0 # default of 0 means no limit
Include only a subset of annotations/labels for each resource type or disable metadata entirely
metadata-collection:
nodes:
annotations: [] # List of annotations to include (explicit or regex)
labels: [] # List of labels to include (explicit or regex)
disable: false # Remove all optional node metadata from the inventory report
namespaces:
annotations: [] # List of annotations to include (explicit or regex)
labels: [] # List of labels to include (explicit or regex)
disable: false # Remove all optional namespace metadata from the inventory report
pods:
annotations: [] # List of annotations to include (explicit or regex)
labels: [] # List of labels to include (explicit or regex)
disable: false # Remove all optional pod metadata from the inventory report
Use this section to configure the Anchore Enterprise API endpoint
anchore:
url: <your anchore api url>
user: <anchore-k8s-inventory_inventory_user>
password: $ANCHORE_K8S_INVENTORY_ANCHORE_PASSWORD
account: <anchore account to send inventory reports>
http:
insecure: true
timeout-seconds: 10
From v1.7.0
, anchore-k8s-inventory will attempt to register as an integration with Enterprise and send health reports
to allow Enterprise to track its status. This requires Enterprise release v5.11.0
or later but the agent will work with
older versions of Enterprise. However, it will only perform the inventory reporting with those Enterprise deployments.
In versions of anchore-k8s-inventory < v0.5.0 the default behavior was to output the inventory report
to stdout every time it was generated. anchore-k8s-inventory v0.5.0 changes this so it will not print
to stdout unless verbose-inventory-reports: true
is set in the config file or
anchore-k8s-inventory is called with the --verbose-inventory-reports
flag.
There are a few configurations that were changed from v0.2.2 to v0.3.0
The request timeout for the kubernetes API was changed from
kubernetes-request-timeout-seconds: 60
to
kubernetes:
request-timeout-seconds: 60
anchore-k8s-inventory will still honor the old configuration. It will prefer the old configuration parameter until it is removed from the config entirely. It is safe to remove the old configuration in favor of the new config.
The namespace configuration was changed from
namespaces:
- all
to
namespace-selectors:
include: []
exclude: []
namespace-selectors
was added to eventually replace namespaces
to allow for both
include and exclude configs. The old namespaces
array will be honored if
namespace-selectors.include
is empty. It is safe to remove namespaces
entirely
in favor of namespace-selectors
Note: This will drop the binary in the ./snapshot/
directory
On Mac
make mac-binary
On Linux
make linux-binary
The Makefile has testing built into it. For unit tests simply run
make unit
To build a docker image, you'll need to provide a kubeconfig.
Note: Docker build requires files to be within the docker build context
docker build -t localhost/anchore-k8s-inventory:latest --build-arg KUBECONFIG=./kubeconfig .
anchore-k8s-inventory comes with shell completion for specifying namespaces, it can be enabled as follows. Run with the --help
command to get the instructions for the shell of your choice
anchore-k8s-inventory completion <zsh|bash|fish>
You can use skaffold for dev. The 'bootstrap-skaffold' make target will clone the chart into the current directory to wire
it up for skaffold to use. To trigger redeployments you'll need to run make linux-binary
and skaffold will rebuild the image
and update the helm release.
make bootstrap-skaffold
make linux-binary
skaffold dev
To create a release of anchore-k8s-inventory, a tag needs to be created that points to a commit in main
that we want to release. This tag shall be a semver prefixed with a v
, e.g. v0.2.7
.
This will trigger a GitHub Action that will create the release.
After the release has been successfully created, make sure to specify the updated version in both Enterprise and the anchore-k8s-inventory Helm Chart in anchore-charts.