All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Fixed an issue where upgrades from older versions (0.1.24) would fail due to incorrect ownership of the File Integrity Operator service accounts. We recommend users upgrade to version 0.1.30 to avoid the issue.
- Restore role bindings for
file-integrity-daemon
to useRole
instead ofClusterRole
. UsingClusterRole
inadvertently broke during upgrades due to expectations by Operator Lifecycle Manager. No action is required to consume this fix besides upgrading to 0.1.29. Please see the bug report for more details.
- The
make release-images
target was updated to publish new catalog images with each release to quay.io/file-integrity-operator/file-integrity-operator-catalog.
- Improved contributor experience by making cleanup
make
targets more robust. - Fixes a bug where the
operator isn't able to install into namespaces other than
openshift-file-integrity
. - Expose a new environment variable used by
make catalog-deploy
to deploy the operator in a non-default namespace. - The operator now uses the appropriate labels to prevent warnings about elevated privileges.
- Minor documentation fixes clarifying details of the release process.
- Removed unused
setup-envtest
dependency required by older versions of Operator SDK. This is no longer needed since upgrading the Operator SDK version. - Removed unnecessary
make
targets for build directories sincebuild/
is tracked as part of the repository. - Increased the grace period for end-to-end tests, giving setup functions enough time to prepare test environments, reducing transient test failures.
- End-to-end tests now use a pod for generating files for test data instead of a daemon set to improve test speeds.
- The File Integrity Operator now explicitly contains an Apache 2.0 License.
- The operator now sets the appropriate namespace for
NodeHasIntegrityFailure
alerts. This helps users understand what component is raising the alert. Please see the corresponding bug report for more details.
- Included
controller-gen
as a dependency using tools.go. - Removed remote naming assumptions from
Makefile
release targets, making the release process more consistent. - Added a
deploy-local
target that deploys the operator using the cluster's image registry. This is primarily for development purposes. - Multiple fixes to help stabilize end-to-end testing.
- Added support for using the File Integrity Operator on s390x and ppc64le architectures.
- Updated operator framework to use operator-sdk version 1.15.0.
- Added formal support for watching all namespaces using
AllNamespaces
in a backwards compatible way to adhere to changes required by the operator-sdk update. - Removed unused test framework code that was obsolete after updating the operator-sdk framework dependency.
- Added additional flexibility to the container build process to support building content across different operating systems.
- Upgraded golang dependency to version 1.17
- Bug 2072058: Use correct data keys for script configMaps
- Update OWNERS file
- deps: Bump the github.com/prometheus/client_golang dependency
- Break apart the make target for release
- e2e: Add the operator logs to artifacts
- Bug 2049206: Re-create AIDE configMaps with missing owner
- Unset --skip-cleanup-error for e2e
- Use golang:1.16 for upstream builder
- Add FileIntegrityConfig MaxBackups
- CSV: We support FIPS
- Move aide.reinit to /run
- Use
oc apply
instead ofoc create
for namespace - Handle IO error in init database more leniently
- Makefile: Make image bundle depend on $TAG
- Fix nil deref in daemonSet upgrade path
- Delete old aide-ds- prefixed daemonSets if they exist
- Makefile: Add test-catalog targets
- Makefile: Apply monitoring resources during deploy-local
- Makefile: Rename IMAGE_FORMAT var
- Use ClusterRole/ClusterRoleBinding for monitoring perms
- Add MCO and CVO related config excludes
- adapt prometheusrule to only alert on currently existing nodes
- Optimize per-node reinit calls
- Handle metrics service during operator start
- Bug 1862022: Per-node reinit during update
- Update Dockerfile.ci goland and base image
- Use fedora-minimal:34 base image
- Move test-specific permissions out of deploy manifests
- Use latest for CSV documentation link
- Enable TLS for controller metrics
- vendor deps
- Add controller-based Prometheus metrics
- Update ignition and MCO dependencies
- Update gosec and fix warnings
- Handle AIDE error code 255
- Update dependencies
- README: Fix instructions to install from OLM
- Add operator and aide-ds pod limits
- daemon: Handle SIGTERM and SIGKILL
- Add an initial CHANGELOG.md and make changelog target
- Exclude the CNI plugin directory
- daemon rewrite and other fixes
- Clean up a test log line
- Add nodeSelector to e2e replacement config
- Remove unused channel for logCollectorMainLoop()
- Compression fixes
- Remove the old AIDE container
- Update CSV links
- Add FileIntegrityStatus and NodeIntegrityStatus events