A Kubernetes Operator to automate deployment of Cryostat.
-
cryostat-core for the core library providing a convenience wrapper and headless stubs for use of JFR using JDK Mission Control internals.
-
cryostat for the main API backend to detect JVMs and manage JFR.
-
cryostat-web for the React graphical frontend included as a submodule in Cryostat and built into Cryostat's OCI images.
-
jfr-datasource for the JFR datasource for Grafana.
-
cryostat-grafana-dashboard for the Grafana dashboard.
kubernetesv1.25+ withOperator Lifecycle Managercert-managerv1.11.5+ (Recommended)
Once deployed, the cryostat instance can be accessed via web browser
at the URL provided by:
kubectl get cryostat -o jsonpath='{$.items[0].status.applicationUrl}'
To use Cryostat to monitor or profile Cryostat itself - since it is also an available JVM target -
you may use the Cryostat web UI to define a Custom Target with the connection URL localhost:0.
This is a special value which tells Cryostat's JVM that it should connect to itself directly, without
the need to expose a JMX port over the network.
The operator's primary installation method is via OperatorHub. A more detailed installation guide is available here. Installation of the Operator without OLM/OperatorHub is intended for development purposes and is not a supported release configuration.
The operator can be deployed using OLM using make deploy_bundle. This will
deploy quay.io/cryostat/cryostat-operator-bundle:$IMAGE_VERSION to
your configured cluster using oc or kubectl (kubeconfig). You can set the
variables IMAGE_NAMESPACE or IMAGE_VERSION to deploy different builds of
the bundle. Once this is complete, the Cryostat Operator will be deployed
and running in your cluster.
make install will create CustomResourceDefinitions and do other setup
required to prepare the cluster for deploying the operator, using oc or
kubectl on whichever OpenShift/Kubernetes cluster is configured with the local client.
make uninstall destroys the CRDs and undoes the setup.
make run can be used to run the operator controller manager as a process on
your local development machine and observing/interacting with your cluster.
This may be useful in some development scenarios, however in this case the
operator process will not have access to certain in-cluster resources such as
environment variables or service account token files.
make deploy will deploy the operator using static manifests to an arbitrary namespace (default to cryostat-operator-system).
make DEPLOY_NAMESPACE=foo-namespace deploycan be used to deploy to a namespace namedfoo-namespace. For a convenient shorthand, usemake DEPLOY_NAMESPACE=$(kubectl config view --minify -o 'jsonpath={.contexts[0].context.namespace}') deployto deploy to the currently active OpenShift/Kubernetes namespace.make undeploywill likewise remove the operator, and also uses theDEPLOY_NAMESPACEvariable. This also respects theIMAGE_TAGenvironment variable, so that different versions of the operator can be easily deployed.- To obtain such static manifests remotely without further customizations, use:
# Replace ref with any version tag or branch kubectl kustomize "https://github.com/cryostatio/cryostat-operator.git/config/default?ref=v2.4.0"
Once deployed, the operator deployment will be active in the cluster, but no Cryostat instance will be created. To trigger its creation, add a Cryostat CR using the UI for operator "provided APIs". Full details on the configuration options in the Cryostat CRD can be found in Configuring Cryostat. When running on Kubernetes, see Network Options for additional mandatory configuration in order to access Cryostat outside of the cluster.
For convenience, a full deployment can be created using
kubectl create -f config/samples/operator_v1beta2_cryostat.yaml, or more
simply, make create_cryostat_cr.
The container images used by the operator for the core application,
jfr-datasource, and the Grafana dashboard can be overridden by setting the
RELATED_IMAGE_CORE, RELATED_IMAGE_DATASOURCE, and RELATED_IMAGE_GRAFANA
environment variables, respectively, in the operator deployment.
By default, the operator expects cert-manager to be available in the cluster.
This allows the operator to deploy Cryostat with all communication
between its internal services done over HTTPS. If you wish to disable this
feature and not use cert-manager, you can set the environment variable
DISABLE_SERVICE_TLS=true when you deploy the operator. We provide
make cert_manager and make remove_cert_manager targets to easily
install/remove cert-manager from your cluster.
Users can use oc whoami --show-token to retrieve their OpenShift OAuth token
for the currently logged in user account. This token can be used when directly
interacting with the deployed Cryostat instance(s).
When using the web-client, users can login with their username and password associated with their OpenShift account. User credentials will be remembered for the duration of the session.
If the current user account does not have sufficient permissions to list routes, list endpointslices, or perform other actions that Cryostat requires, then the user may also try to authenticate using the Operator's service account. This, of course, assumes that the user has permission to view this service account's secrets.
oc get secrets | grep cryostat-operator-service-account-token will provide at least one
such operator service account token secret name which can be used - for
example, cryostat-operator-service-account-token-7tt7l. The token can then be retrieved
for use in authenticating as the operator service account:
$ oc describe secret cryostat-operator-service-account-token-7tt7l
Name: cryostat-operator-service-account-token-7tt7l
Namespace: default
Labels: <none>
Annotations: kubernetes.io/created-by: openshift.io/create-dockercfg-secrets
kubernetes.io/service-account.name: cryostat-operator-service-account
kubernetes.io/service-account.uid: 692aa8c7-081e-4a51-9355-be3eaa8f9fa6
Type: kubernetes.io/service-account-token
Data
====
ca.crt: 7209 bytes
namespace: 7 bytes
service-ca.crt: 8422 bytes
token: eyJhbGciOiJSUzI1NiIsImtpZCI6IkhYZC13eDdGVGwyQzdGNVpZVndScEZ2VmRxWTlzbnBUUG9HRkJpejJkV3cifQ.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.M7C1V0bN3aILBflO7TTOTikw7wLGRJ79-OkCDQIZbu71QLdX05jyCxxtlH32lr8jz6HwxfXXweh3ifG_2lbe7_TbM8jxmBoMdLuc4Q_akpmA-GQuDPrRxfHGJApYGQ6CVug3KHSrQwj2M4QrSUz7FoeQGaOH9BnWj1TrHGmOZUPJ6u7JSu2OwoLBda6rF-M4Bl72DmkyMAzikreRgPEk4D7gTCY0yNvsQDuUAwpFwmEukRC2WyTAVTpKPgThZUk-UJ-dXufbhAcqIRt6jeCQ19_Bo0zXc_ELgQydxuTack1ndT3HwRmwwNuZDFv-G3Y0YdjfRh00DqEvSn9ynZzwueDCJUxlHdznytfUWk9PA712JENpFC7b-zSHnjymIcFeUd8s_Zq_-JKrDIPnH0oZDRO_MUpKEC7Jz_8SeFJHLLGfBZt_aP4VwQHEUThiFQPwrfbd8tppUG2TKcekPScKcauy-BCI52odBzapP6meilMQVrmRtu7i30L05vgQiST_OsmSP8CuKW13a-leCCtN_aNQGqlWvLhP81H95ui-PvMzwMIDlfDZ03ycuYg4R4eUG3nUq7-42wrSdFLo8gm9wsl7y1ZRMQwHR1DCVBbHYS0iFOcmwto2Ejlrgvn3Cs0pDS7pDVoFkH2FsTopEw3jXtnkMs15mSmBnHz-UjF-l08
or more briefly:
$ oc get -o jsonpath='{.data.token}' secret cryostat-operator-service-account-token-7tt7l | base64 -d
eyJhbGciOiJSUzI1NiIsImtpZCI6IkhYZC13eDdGVGwyQzdGNVpZVndScEZ2VmRxWTlzbnBUUG9HRkJpejJkV3cifQ.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.M7C1V0bN3aILBflO7TTOTikw7wLGRJ79-OkCDQIZbu71QLdX05jyCxxtlH32lr8jz6HwxfXXweh3ifG_2lbe7_TbM8jxmBoMdLuc4Q_akpmA-GQuDPrRxfHGJApYGQ6CVug3KHSrQwj2M4QrSUz7FoeQGaOH9BnWj1TrHGmOZUPJ6u7JSu2OwoLBda6rF-M4Bl72DmkyMAzikreRgPEk4D7gTCY0yNvsQDuUAwpFwmEukRC2WyTAVTpKPgThZUk-UJ-dXufbhAcqIRt6jeCQ19_Bo0zXc_ELgQydxuTack1ndT3HwRmwwNuZDFv-G3Y0YdjfRh00DqEvSn9ynZzwueDCJUxlHdznytfUWk9PA712JENpFC7b-zSHnjymIcFeUd8s_Zq_-JKrDIPnH0oZDRO_MUpKEC7Jz_8SeFJHLLGfBZt_aP4VwQHEUThiFQPwrfbd8tppUG2TKcekPScKcauy-BCI52odBzapP6meilMQVrmRtu7i30L05vgQiST_OsmSP8CuKW13a-leCCtN_aNQGqlWvLhP81H95ui-PvMzwMIDlfDZ03ycuYg4R4eUG3nUq7-42wrSdFLo8gm9wsl7y1ZRMQwHR1DCVBbHYS0iFOcmwto2Ejlrgvn3Cs0pDS7pDVoFkH2FsTopEw3jXtnkMs15mSmBnHz-UjF-l08
go(seego.modversion)operator-sdkv1.42.0podmanordockerjqv1.6+yqv4.35+ginkgo(Optional)
make generate manifests manager will trigger code/YAML generation and compile
the operator controller manager, along with running some code quality checks.
make oci-build will build an OCI image from the generated YAML and compiled
binary to the local registry, tagged as
quay.io/crystatio/cryostat-operator. This tag can be overridden by
setting the environment variables IMAGE_NAMESPACE and OPERATOR_NAME.
IMAGE_VERSION can also be set to override the tagged version.
make bundle will create an OLM bundle. This will generate a CSV, CRDs and
other manifests, and other required configurations for an OLM bundle versioned
with version $IMAGE_VERSION in the bundle/ directory. make bundle-build
will create an OCI image of this bundle, which can then be pushed to an image
repository such as quay.io.
make catalog-build will build an OCI image of the operator catalog (i.e. index)
with version $IMAGE_VERSION that includes the bundle image of the same version.
An invocation like
export IMAGE_NAMESPACE=quay.io/some-user export IMAGE_VERSION=test-version
make generate manifests manager oci-build bundle bundle-build
podman image prune -f && podman push $IMAGE_NAMESPACE/cryostat-operator:$IMAGE_VERSION && podman push $IMAGE_NAMESPACE/cryostat-operator-bundle:$IMAGE_VERSION
make deploy_bundle
is handy for local development testing using ex. CodeReady Containers. This
exercises a similar build and deployment path as what end users using OLM and
OperatorHub will eventually receive.
- (optional) oc
- (optional) crc
- (for
test-scorecard-local)dockerorpodman - (optional, for
test-scorecard-local-kind) kind anddockerorpodman
make test-envtest will run controller tests using ginkgo if installed, or go test if
not, requiring no cluster connection.
make test-scorecard will run the Operator SDK's scorecard test suite. This requires a
Kubernetes or OpenShift cluster to be available and logged in with your kubectl or oc
client. The recommended setup for development testing is CodeReady Containers (crc).
make test-scorecard-local runs the custom scorecard tests from the local source
against BUNDLE_IMG. It uses IMAGE_BUILDER (podman by default, or docker) to
pull the bundle and copy its manifests and metadata into a temporary directory.
The container and temporary directory are removed when the tests finish. There is
no need to regenerate ./bundle when switching bundle images. For example:
BUNDLE_IMG=quay.io/cryostat/cryostat-operator-bundle:latest-ocp \
SCORECARD_TEST_SELECTION=console-plugin make test-scorecard-localWith SCORECARD_TEST_ONLY=true, set BUNDLE_IMG to the bundle already installed
in SCORECARD_NAMESPACE.
Before the scorecard tests are run, all cryostat and cryostat-operator resources will be deleted to ensure a clean slate.
make test will run all tests.
cryostat-agent deploys the sample Java application with a bundled Cryostat Agent,
configures mutual TLS using the Cryostat CA, waits for Agent registration, and runs
the recording lifecycle used by cryostat-recording. cryostat-agent-autoconfig
deploys an image without the Agent and labels only the Deployment. It verifies
that the Deployment webhook propagates the labels and the Pod webhook injects the
Agent while preserving the application's JVM options, then checks registration
and runs the same recording lifecycle. Both tests call Cryostat through its
authenticated in-cluster Service. Neither application exposes remote JMX.
Application, init-container, and Cryostat logs are included in the test output.
Sample resources are owned by the test's Cryostat CR and removed with it.
The sample source and image builds live in
cryostatio/test-applications.
Its existing workflows build both samples and publish them under
quay.io/redhat-java-monitoring:
cryostat-operator-scorecard-app: application without an Agent.cryostat-operator-scorecard-agent-app: application with the Agent JAR at/opt/cryostat/cryostat-agent-shaded.jar.
make scorecard-build builds only the test runner, which stays under
quay.io/cryostat. The runner uses the published sample images by default.
SCORECARD_SAMPLE_VERSION selects their tag (default latest); override
SCORECARD_SAMPLE_IMG and SCORECARD_AGENT_IMG to embed specific references,
including digests, in the runner. The sample tag is independent of the runner's
timestamped tag. For tests run directly from source, use
CRYOSTAT_SCORECARD_SAMPLE_IMAGE and CRYOSTAT_SCORECARD_AGENT_IMAGE.
To test unpublished sample changes, supply a local test-applications checkout.
The Kind flow then builds both samples using that checkout's scripts and serves
them from its local registry:
SCORECARD_APPLICATIONS_DIR=/path/to/test-applications \
OPERATOR_IMG=quay.io/cryostat/cryostat-operator:latest \
CORE_IMG=quay.io/cryostat/cryostat:latest \
SCORECARD_TEST_SELECTION="cryostat-agent,cryostat-agent-autoconfig" \
make test-scorecard-local-kindThe sample scripts require Podman. They copy the bundled JAR from
AGENT_INIT_IMG, matching the Operator's injection artifact. To build samples
separately for an existing cluster:
MANIFEST_PUSH=false make scorecard-app-build \
SCORECARD_APPLICATIONS_DIR=/path/to/test-applications \
PLATFORMS=linux/amd64 \
SCORECARD_SAMPLE_IMG=quay.io/me/scorecard-app:test \
SCORECARD_AGENT_IMG=quay.io/me/scorecard-agent-app:test
# Make these images available to the cluster before running the tests.
CRYOSTAT_SCORECARD_SAMPLE_IMAGE=quay.io/me/scorecard-app:test \
CRYOSTAT_SCORECARD_AGENT_IMAGE=quay.io/me/scorecard-agent-app:test \
SCORECARD_TEST_SELECTION="cryostat-agent,cryostat-agent-autoconfig" \
make test-scorecard-localUse the Go version in go.mod and the Operator SDK version in Makefile
(bin/operator-sdk takes precedence over a globally installed SDK). From the
Operator repository root, build the runner with the published sample references:
export CUSTOM_SCORECARD_VERSION="5.0.0-$(date -u +%Y%m%d%H%M%S)"
export CUSTOM_SCORECARD_IMG="quay.io/cryostat/cryostat-operator-scorecard:${CUSTOM_SCORECARD_VERSION}"
# Prefer a published version/date tag or digest when preparing a runner release.
export SCORECARD_SAMPLE_IMG="quay.io/redhat-java-monitoring/cryostat-operator-scorecard-app:<published-tag>"
export SCORECARD_AGENT_IMG="quay.io/redhat-java-monitoring/cryostat-operator-scorecard-agent-app:<published-tag>"
make scorecard-build IMAGE_BUILDER=podman \
PLATFORMS=linux/amd64,linux/arm64 MANIFEST_PUSH=false
make bundleThe build embeds the sample image references in the runner. make bundle
regenerates bundle/tests/scorecard/config.yaml from the updated Scorecard
configuration; it does not publish an image. Review the generated diff, including
that file and config/scorecard/patches/custom.config.yaml, before including the
updated Scorecard configuration in the PR.
For an AMD64-only test cluster, PLATFORMS=linux/amd64 also works.
To make the runner available from Quay before merge, optionally publish it after the build and bundle regeneration succeed:
podman login quay.io
podman manifest push --all "$CUSTOM_SCORECARD_IMG" "docker://$CUSTOM_SCORECARD_IMG"This command uploads a container image. Git commits and pushes are separate.
Once the Operator change is merged upstream, its publishing workflow builds and
publishes the checked-in runner tag under quay.io/cryostat if it does not already
exist. Before merge, use the local Kind flow to validate the runner and replace
the build-number placeholders in both Scorecard configs with a fresh timestamped
tag.
make test-scorecard-local-kind runs the scorecard suite end-to-end against a
dedicated kind cluster, with no external image registry
required. It is a turnkey wrapper that:
- creates (or reuses) a kind cluster named
cryostat-scorecard, configured with a containerd registry mirror; - starts a local container registry, published to the host as
localhost:5000and mirrored to the same reference in each node's containerd, so one image reference resolves both on the host and in the cluster; - installs the cluster prerequisites: OLM, cert-manager, and ingress-nginx (with an in-cluster CoreDNS rewrite so the tests can reach Cryostat via its ingress);
- builds the scorecard test image and the operator bundle locally, pushes them to the
local registry, and runs the scorecard suite against the on-disk
./bundle.
Because the scorecard test image and bundle are served from the local registry, you can
validate that a set of operand images passes the scorecard suite without publishing
any scorecard-specific images. Supply the operand images under test via OPERATOR_IMG
and CORE_IMG (and any other RELATED_IMAGE_* overrides), for example:
CORE_IMG=quay.io/my-user/cryostat:my-feature-pr \
OPERATOR_IMG=quay.io/my-user/cryostat-operator:my-feature-pr \
make test-scorecard-local-kindRun a subset of tests with SCORECARD_TEST_SELECTION (e.g.
SCORECARD_TEST_SELECTION=cryostat-cr), and tear everything down afterwards with
make clean-scorecard-local-kind.
The container runtime backing kind is auto-detected from IMAGE_BUILDER (podman by
default); set KIND_RUNTIME=docker if kind uses the docker provider. Under the rootless
podman provider, run this as the only kind cluster on the host — creating it alongside
another running kind cluster can disrupt cluster networking. Nothing here needs root.
The scorecard-kind-prereqs and scorecard-kind-push targets are shared with the CI
scorecard job, and all of these can be invoked individually against an existing cluster:
scorecard-kind-cluster, scorecard-kind-registry, scorecard-kind-prereqs,
scorecard-kind-push (see make help). CI does not use scorecard-kind-registry:
helm/kind-action's built-in registry support provides the equivalent registry there,
addressed as kind-registry:5000 via the hosts entry the action adds on the runner.