k8s-image-availability-exporter (or k8s-iae for short) is a Prometheus exporter that warns you proactively about images that are defined in Kubernetes objects (e.g., an image
field in the Deployment) but are not available in the container registry (such as Docker Registry, etc.).
Receiving alerts when container images related to running Kubernetes controllers are missing helps you to solve the problem before it manifests itself. For more details on the reasons behind k8s-iae and how it works, please read this article.
- Deploying / Installing k8s-iae in your Kubernetes cluster
- Prometheus integration to scrape metrics
- Alerting based on k8s-iae metrics
- Configuration
- Metrics for Prometheus provided by k8s-iae
- Compatibility
Ready-to-use container images are available in the Deckhouse registry:
docker pull registry.deckhouse.io/k8s-image-availability-exporter/k8s-image-availability-exporter:latest
The helm chart is available on artifacthub. Follow instructions on the page to install it.
Here's how you can configure Prometheus or prometheus-operator to scrape metrics from k8s-image-availability-exporter
.
- job_name: image-availability-exporter
honor_labels: true
metrics_path: '/metrics'
scheme: http
kubernetes_sd_configs:
- role: pod
namespaces:
names:
- kube-system
relabel_configs:
- source_labels: [__meta_kubernetes_pod_label_app]
regex: image-availability-exporter
action: keep
apiVersion: monitoring.coreos.com/v1
kind: PodMonitor
metadata:
name: image-availability-exporter
namespace: kube-system
spec:
podMetricsEndpoints:
- port: http-metrics
scheme: http
honorLabels: true
scrapeTimeout: 10s
selector:
matchLabels:
app: image-availability-exporter
namespaceSelector:
matchNames:
- kube-system
Here's how to alert based on these metrics:
groups:
- alert: DeploymentImageUnavailable
expr: |
max by (namespace, name, container, image) (
k8s_image_availability_exporter_available{kind="deployment"} == 0
)
annotations:
message: >
Image {{`{{ $labels.image }}`}} from container {{`{{ $labels.container }}`}}
in deployment {{`{{ $labels.name }}`}}
from namespace {{`{{ $labels.namespace }}`}}
is not available in docker registry.
labels:
severity: critical
- alert: StatefulSetImageUnavailable
expr: |
max by (namespace, name, container, image) (
k8s_image_availability_exporter_available{kind="statefulset"} == 0
)
annotations:
message: >
Image {{`{{ $labels.image }}`}} from container {{`{{ $labels.container }}`}}
in statefulSet {{`{{ $labels.name }}`}}
from namespace {{`{{ $labels.namespace }}`}}
is not available in docker registry.
labels:
severity: critical
- alert: DaemonSetImageUnavailable
expr: |
max by (namespace, name, container, image) (
k8s_image_availability_exporter_available{kind="daemonset"} == 0
)
annotations:
message: >
Image {{`{{ $labels.image }}`}} from container {{`{{ $labels.container }}`}}
in daemonSet {{`{{ $labels.name }}`}}
from namespace {{`{{ $labels.namespace }}`}}
is not available in docker registry.
labels:
severity: critical
- alert: CronJobImageUnavailable
expr: |
max by (namespace, name, container, image) (
k8s_image_availability_exporter_available{kind="cronjob"} == 0
)
annotations:
message: >
Image {{`{{ $labels.image }}`}} from container {{`{{ $labels.container }}`}}
in cronJob {{`{{ $labels.name }}`}}
from namespace {{`{{ $labels.namespace }}`}}
is not available in docker registry.
labels:
severity: critical
Usage of k8s-image-availability-exporter:
-allow-plain-http
whether to fallback to HTTP scheme for registries that don't support HTTPS
-bind-address string
address:port to bind /metrics endpoint to (default ":8080")
-capath value
path to a file that contains CA certificates in the PEM format
-check-interval duration
image re-check interval (default 1m0s)
-default-registry string
default registry to use in absence of a fully qualified image name, defaults to "index.docker.io"
-force-check-disabled-controllers value
comma-separated list of controller kinds for which image is forcibly checked, even when workloads are disabled or suspended. Acceptable values include "Deployment", "StatefulSet", "DaemonSet", "Cronjob" or "*" for all kinds (this option is case-insensitive)
-ignored-images string
tilde-separated image regexes to ignore, each image will be checked against this list of regexes
-image-mirror value
Add a mirror repository (format: original=mirror)
-namespace-label string
namespace label for checks
-skip-registry-cert-verification
whether to skip registries' certificate verification
The following metrics for Prometheus are provided:
k8s_image_availability_exporter_available
— non-zero indicates successful image check.k8s_image_availability_exporter_absent
— non-zero indicates an image's manifest absence from container registry.k8s_image_availability_exporter_bad_image_format
— non-zero indicates incorrectimage
field format.k8s_image_availability_exporter_registry_unavailable
— non-zero indicates general registry unavailiability, perhaps, due to network outage.k8s_image_availability_exporter_authentication_failure
— non-zero indicates authentication error to container registry, verify imagePullSecrets.k8s_image_availability_exporter_authorization_failure
— non-zero indicates authorization error to container registry, verify imagePullSecrets.k8s_image_availability_exporter_unknown_error
— non-zero indicates an error that failed to be classified, consult exporter's logs for additional information.
Each metric has the following labels:
namespace
- namespace namecontainer
- container nameimage
- image URL in the registrykind
- Kubernetes controller kind, namelydeployment
,statefulset
,daemonset
orcronjob
name
- controller name
k8s-image-availability-exporter is compatible with Kubernetes 1.15+ and Docker Registry V2 compliant container registries.
Since the exporter operates as a Deployment, it does not support container registries that should be accessed via authorization on a node.