Compliance Operator release notes

The Compliance Operator is now stable and the release channel is upgraded to stable. Future releases will follow Semantic Versioning. To access the latest release, see Updating the Compliance Operator.

Before this update, the compliance_operator_compliance_scan_error_total metric had an ERROR label with a different value for each error message. With this update, the compliance_operator_compliance_scan_error_total metric does not increase in values. (OCPBUGS-1803)

Before this update, the ocp4-api-server-audit-log-maxsize rule would result in a FAIL state. With this update, the error message has been removed from the metric, decreasing the cardinality of the metric in line with best practices. (OCPBUGS-7520)

Before this update, the rhcos4-enable-fips-mode rule description was misleading that fips could be enabled after installation. With this update, the rhcos4-enable-fips-mode rule description clarifies that fips must be enabled at install time. (OCPBUGS-8358)

The Compliance Operator now supports timeout configuration for Scanner Pods. The timeout is specified in the ScanSetting object. If the scan is not completed within the timeout, the scan retries until the maximum number of retries is reached. See Configuring ScanSetting timeout for more information.

Before this update, Compliance Operator remediations required variables as inputs. Remediations without variables set were applied cluster-wide and resulted in stuck nodes, even though it appeared the remediation applied correctly. With this update, the Compliance Operator validates if a variable needs to be supplied using a TailoredProfile for a remediation. (OCPBUGS-3864)

Before this update, the instructions for ocp4-kubelet-configure-tls-cipher-suites were incomplete, requiring users to refine the query manually. With this update, the query provided in ocp4-kubelet-configure-tls-cipher-suites returns the actual results to perform the audit steps. (OCPBUGS-3017)

Before this update,ScanSettingBinding objects created without a settingRef variable did not use an appropriate default value. With this update, the ScanSettingBinding objects without a settingRef variable use the default value. (OCPBUGS-3420)

Before this update, system reserved parameters were not generated in kubelet configuration files, causing the Compliance Operator to fail to unpause the machine config pool. With this update, the Compliance Operator omits system reserved parameters during machine configuration pool evaluation. (OCPBUGS-4445)

Before this update, ComplianceCheckResult objects did not have correct descriptions. With this update, the Compliance Operator sources the ComplianceCheckResult information from the rule description. (OCPBUGS-4615)

Before this update, the Compliance Operator did not check for empty kubelet configuration files when parsing machine configurations. As a result, the Compliance Operator would panic and crash. With this update, the Compliance Operator implements improved checking of the kubelet configuration data structure and only continues if it is fully rendered. (OCPBUGS-4621)

Before this update, the Compliance Operator generated remediations for kubelet evictions based on machine config pool name and a grace period, resulting in multiple remediations for a single eviction rule. With this update, the Compliance Operator applies all remediations for a single rule. (OCPBUGS-4338)

Before this update, re-running scans on remediations that previously Applied might have been marked as Outdated after rescans were performed, despite no changes in the remediation content. The comparison of scans did not account for remediation metadata correctly. With this update, remediations retain the previously generated Applied status. (OCPBUGS-6710)

Before this update, a regression occurred when attempting to create a ScanSettingBinding that was using a TailoredProfile with a non-default MachineConfigPool marked the ScanSettingBinding as Failed. With this update, functionality is restored and custom ScanSettingBinding using a TailoredProfile performs correctly. (OCPBUGS-6827)

Before this update, some kubelet configuration parameters did not have default values. With this update, the following parameters contain default values (OCPBUGS-6708):

Before this update, the selinux_confinement_of_daemons rule failed running on the kubelet because of the permissions necessary for the kubelet to run. With this update, the selinux_confinement_of_daemons rule is disabled. (OCPBUGS-6968)

The Compliance Operator now supports Payment Card Industry Data Security Standard (PCI-DSS) ocp4-pci-dss and ocp4-pci-dss-node profiles on the ppc64le architecture.

Previously, the Compliance Operator did not support the Payment Card Industry Data Security Standard (PCI DSS) ocp4-pci-dss and ocp4-pci-dss-node profiles on different architectures such as ppc64le. Now, the Compliance Operator supports ocp4-pci-dss and ocp4-pci-dss-node profiles on the ppc64le architecture. (OCPBUGS-3252)

Previously, after the recent update to version 0.1.57, the rerunner service account (SA) was no longer owned by the cluster service version (CSV), which caused the SA to be removed during the Operator upgrade. Now, the CSV owns the rerunner SA in 0.1.59, and upgrades from any previous version will not result in a missing SA. (OCPBUGS-3452)

In 0.1.57, the Operator started the controller metrics endpoint listening on port 8080. This resulted in TargetDown alerts since cluster monitoring expected port is 8383. With 0.1.59, the Operator starts the endpoint listening on port 8383 as expected. (OCPBUGS-3097)

KubeletConfig checks changed from Node to Platform type. KubeletConfig checks the default configuration of the KubeletConfig. The configuration files are aggregated from all nodes into a single location per node pool. See Evaluating KubeletConfig rules against default configuration values.

The ScanSetting Custom Resource now allows users to override the default CPU and memory limits of scanner pods through the scanLimits attribute. For more information, see Increasing Compliance Operator resource limits.

A PriorityClass object can now be set through ScanSetting. This ensures the Compliance Operator is prioritized and minimizes the chance that the cluster falls out of compliance. For more information, see Setting PriorityClass for ScanSetting scans.

Previously, the Compliance Operator hard-coded notifications to the default openshift-compliance namespace. If the Operator were installed in a non-default namespace, the notifications would not work as expected. Now, notifications work in non-default openshift-compliance namespaces. (BZ#2060726)

Previously, the Compliance Operator was unable to evaluate default configurations used by kubelet objects, resulting in inaccurate results and false positives. This new feature evaluates the kubelet configuration and now reports accurately. (BZ#2075041)

Previously, the Compliance Operator reported the ocp4-kubelet-configure-event-creation rule in a FAIL state after applying an automatic remediation because the eventRecordQPS value was set higher than the default value. Now, the ocp4-kubelet-configure-event-creation rule remediation sets the default value, and the rule applies correctly. (BZ#2082416)

The ocp4-configure-network-policies rule requires manual intervention to perform effectively. New descriptive instructions and rule updates increase applicability of the ocp4-configure-network-policies rule for clusters using Calico CNIs. (BZ#2091794)

Previously, the Compliance Operator would not clean up pods used to scan infrastructure when using the debug=true option in the scan settings. This caused pods to be left on the cluster even after deleting the ScanSettingBinding. Now, pods are always deleted when a ScanSettingBinding is deleted.(BZ#2092913)

Previously, the Compliance Operator used an older version of the operator-sdk command that caused alerts about deprecated functionality. Now, an updated version of the operator-sdk command is included and there are no more alerts for deprecated functionality. (BZ#2098581)

Previously, the Compliance Operator would fail to apply remediations if it could not determine the relationship between kubelet and machine configurations. Now, the Compliance Operator has improved handling of the machine configurations and is able to determine if a kubelet configuration is a subset of a machine configuration. (BZ#2102511)

Previously, the rule for ocp4-cis-node-master-kubelet-enable-cert-rotation did not properly describe success criteria. As a result, the requirements for RotateKubeletClientCertificate were unclear. Now, the rule for ocp4-cis-node-master-kubelet-enable-cert-rotation reports accurately regardless of the configuration present in the kubelet configuration file. (BZ#2105153)

Previously, the rule for checking idle streaming timeouts did not consider default values, resulting in inaccurate rule reporting. Now, more robust checks ensure increased accuracy in results based on default configuration values. (BZ#2105878)

Previously, the Compliance Operator would fail to fetch API resources when parsing machine configurations without Ignition specifications, which caused the api-check-pods processes to crash loop. Now, the Compliance Operator handles Machine Config Pools that do not have Ignition specifications correctly. (BZ#2117268)

Previously, rules evaluating the modprobe configuration would fail even after applying remediations due to a mismatch in values for the modprobe configuration. Now, the same values are used for the modprobe configuration in checks and remediations, ensuring consistent results. (BZ#2117747)

Specifying Install into all namespaces in the cluster or setting the WATCH_NAMESPACES environment variable to "" no longer affects all namespaces. Any API resources installed in namespaces not specified at the time of Compliance Operator installation is no longer be operational. API resources might require creation in the selected namespace, or the openshift-compliance namespace by default. This change improves the Compliance Operator’s memory usage.

Previously, the ocp4-kubelet-enable-streaming-connections rule contained an incorrect variable comparison, resulting in false positive scan results. Now, the Compliance Operator provides accurate scan results when setting streamingConnectionIdleTimeout. (BZ#2069891)

Previously, group ownership for /etc/openvswitch/conf.db was incorrect on IBM Z architectures, resulting in ocp4-cis-node-worker-file-groupowner-ovs-conf-db check failures. Now, the check is marked NOT-APPLICABLE on IBM Z architecture systems. (BZ#2072597)

Previously, the ocp4-cis-scc-limit-container-allowed-capabilities rule reported in a FAIL state due to incomplete data regarding the security context constraints (SCC) rules in the deployment. Now, the result is MANUAL, which is consistent with other checks that require human intervention. (BZ#2077916)

Previously, the following rules failed to account for additional configuration paths for API servers and TLS certificates and keys, resulting in reported failures even if the certificates and keys were set properly:

Previously, the content_rule_oauth_or_oauthclient_inactivity_timeout rule did not account for a configurable timeout set by the deployment when assessing compliance for timeouts. This resulted in the rule failing even if the timeout was valid. Now, the Compliance Operator uses the var_oauth_inactivity_timeout variable to set valid timeout length. (BZ#2081952)

Previously, the Compliance Operator used administrative permissions on namespaces not labeled appropriately for privileged use, resulting in warning messages regarding pod security-level violations. Now, the Compliance Operator has appropriate namespace labels and permission adjustments to access results without violating permissions. (BZ#2088202)

Previously, applying auto remediations for rhcos4-high-master-sysctl-kernel-yama-ptrace-scope and rhcos4-sysctl-kernel-core-pattern resulted in subsequent failures of those rules in scan results, even though they were remediated. Now, the rules report PASS accurately, even after remediations are applied.(BZ#2094382)

Previously, the Compliance Operator would fail in a CrashLoopBackoff state because of out-of-memory exceptions. Now, the Compliance Operator is improved to handle large machine configuration data sets in memory and function correctly. (BZ#2094854)

When "debug":true is set within the ScanSettingBinding object, the pods generated by the ScanSettingBinding object are not removed when that binding is deleted. As a workaround, run the following command to delete the remaining pods:

$ oc delete pods -l compliance.openshift.io/scan-name=ocp4-cis

The FedRAMP high SCAP profile is now available for use in OpenShift Container Platform environments. For more information, See Supported compliance profiles.

Previously, the OpenScap container would crash due to a mount permission issue in a security environment where DAC_OVERRIDE capability is dropped. Now, executable mount permissions are applied to all users. (BZ#2082151)

Previously, the compliance rule ocp4-configure-network-policies could be configured as MANUAL. Now, compliance rule ocp4-configure-network-policies is set to AUTOMATIC. (BZ#2072431)

Previously, the Cluster Autoscaler would fail to scale down because the Compliance Operator scan pods were never removed after a scan. Now, the pods are removed from each node by default unless explicitly saved for debugging purposes. (BZ#2075029)

Previously, applying the Compliance Operator to the KubeletConfig would result in the node going into a NotReady state due to unpausing the Machine Config Pools too early. Now, the Machine Config Pools are unpaused appropriately and the node operates correctly. (BZ#2071854)

Previously, the Machine Config Operator used base64 instead of url-encoded code in the latest release, causing Compliance Operator remediation to fail. Now, the Compliance Operator checks encoding to handle both base64 and url-encoded Machine Config code and the remediation applies correctly. (BZ#2082431)

When "debug":true is set within the ScanSettingBinding object, the pods generated by the ScanSettingBinding object are not removed when that binding is deleted. As a workaround, run the following command to delete the remaining pods:

$ oc delete pods -l compliance.openshift.io/scan-name=ocp4-cis

Previously, the openshift-compliance content did not include platform-specific checks for network types. As a result, OVN- and SDN-specific checks would show as failed instead of not-applicable based on the network configuration. Now, new rules contain platform checks for networking rules, resulting in a more accurate assessment of network-specific checks. (BZ#1994609)

Previously, the ocp4-moderate-routes-protected-by-tls rule incorrectly checked TLS settings that results in the rule failing the check, even if the connection secure SSL/TLS protocol. Now, the check properly evaluates TLS settings that are consistent with the networking guidance and profile recommendations. (BZ#2002695)

Previously, ocp-cis-configure-network-policies-namespace used pagination when requesting namespaces. This caused the rule to fail because the deployments truncated lists of more than 500 namespaces. Now, the entire namespace list is requested, and the rule for checking configured network policies works for deployments with more than 500 namespaces. (BZ#2038909)

Previously, remediations using the sshd jinja macros were hard-coded to specific sshd configurations. As a result, the configurations were inconsistent with the content the rules were checking for and the check would fail. Now, the sshd configuration is parameterized and the rules apply successfully. (BZ#2049141)

Previously, the ocp4-cluster-version-operator-verify-integrity always checked the first entry in the Cluter Version Operator (CVO) history. As a result, the upgrade would fail in situations where subsequent versions of {product-name} would be verified. Now, the compliance check result for ocp4-cluster-version-operator-verify-integrity is able to detect verified versions and is accurate with the CVO history. (BZ#2053602)

Previously, the ocp4-api-server-no-adm-ctrl-plugins-disabled rule did not check for a list of empty admission controller plugins. As a result, the rule would always fail, even if all admission plugins were enabled. Now, more robust checking of the ocp4-api-server-no-adm-ctrl-plugins-disabled rule accurately passes with all admission controller plugins enabled. (BZ#2058631)

Previously, scans did not contain platform checks for running against Linux worker nodes. As a result, running scans against worker nodes that were not Linux-based resulted in a never ending scan loop. Now, the scan schedules appropriately based on platform type and labels complete successfully. (BZ#2056911)

Previously, some rules associated with extended Open Vulnerability and Assessment Language (OVAL) definitions had a checkType of None. This was because the Compliance Operator was not processing extended OVAL definitions when parsing rules. With this update, content from extended OVAL definitions is parsed so that these rules now have a checkType of either Node or Platform. (BZ#2040282)

Previously, a manually created MachineConfig object for KubeletConfig prevented a KubeletConfig object from being generated for remediation, leaving the remediation in the Pending state. With this release, a KubeletConfig object is created by the remediation, regardless if there is a manually created MachineConfig object for KubeletConfig. As a result, KubeletConfig remediations now work as expected. (BZ#2040401)

The Compliance Operator now supports the following compliance benchmarks for the Payment Card Industry Data Security Standard (PCI DSS):

Additional rules and remediations for FedRAMP moderate impact level are added to the OCP4-moderate, OCP4-moderate-node, and rhcos4-moderate profiles.

Remediations for KubeletConfig are now available in node-level profiles.

Previously, if your cluster was running OpenShift Container Platform 4.6 or earlier, remediations for USBGuard-related rules would fail for the moderate profile. This is because the remediations created by the Compliance Operator were based on an older version of USBGuard that did not support drop-in directories. Now, invalid remediations for USBGuard-related rules are not created for clusters running OpenShift Container Platform 4.6. If your cluster is using OpenShift Container Platform 4.6, you must manually create remediations for USBGuard-related rules.

Previously, when rendering remediations, the compliance operator would check that the remediation was well-formed by using a regular expression that was too strict. As a result, some remediations, such as those that render sshd_config, would not pass the regular expression check and therefore, were not created. The regular expression was found to be unnecessary and removed. Remediations now render correctly. (BZ#2033009)

In this release, the strictNodeScan option is now added to the ComplianceScan, ComplianceSuite and ScanSetting CRs. This option defaults to true which matches the previous behavior, where an error occurred if a scan was not able to be scheduled on a node. Setting the option to false allows the Compliance Operator to be more permissive about scheduling scans. Environments with ephemeral nodes can set the strictNodeScan value to false, which allows a compliance scan to proceed, even if some of the nodes in the cluster are not available for scheduling.

You can now customize the node that is used to schedule the result server workload by configuring the nodeSelector and tolerations attributes of the ScanSetting object. These attributes are used to place the ResultServer pod, the pod that is used to mount a PV storage volume and store the raw Asset Reporting Format (ARF) results. Previously, the nodeSelector and the tolerations parameters defaulted to selecting one of the control plane nodes and tolerating the node-role.kubernetes.io/master taint. This did not work in environments where control plane nodes are not permitted to mount PVs. This feature provides a way for you to select the node and tolerate a different taint in those environments.

The Compliance Operator can now remediate KubeletConfig objects.

A comment containing an error message is now added to help content developers differentiate between objects that do not exist in the cluster compared to objects that cannot be fetched.

Rule objects now contain two new attributes, checkType and description. These attributes allow you to determine if the rule pertains to a node check or platform check, and also allow you to review what the rule does.

This enhancement removes the requirement that you have to extend an existing profile to create a tailored profile. This means the extends field in the TailoredProfile CRD is no longer mandatory. You can now select a list of rule objects to create a tailored profile. Note that you must select whether your profile applies to nodes or the platform by setting the compliance.openshift.io/product-type: annotation or by setting the -node suffix for the TailoredProfile CR.

In this release, the Compliance Operator is now able to schedule scans on all nodes irrespective of their taints. Previously, the scan pods would only tolerated the node-role.kubernetes.io/master taint, meaning that they would either ran on nodes with no taints or only on nodes with the node-role.kubernetes.io/master taint. In deployments that use custom taints for their nodes, this resulted in the scans not being scheduled on those nodes. Now, the scan pods tolerate all node taints.

In this release, the Compliance Operator supports the following North American Electric Reliability Corporation (NERC) security profiles:

In this release, the Compliance Operator supports the NIST 800-53 Moderate-Impact Baseline for the Red Hat OpenShift - Node level, ocp4-moderate-node, security profile.

In this release, the remediation template now allows multi-value variables.

With this update, the Compliance Operator can change remediations based on variables that are set in the compliance profile. This is useful for remediations that include deployment-specific values such as time outs, NTP server host names, or similar. Additionally, the ComplianceCheckResult objects now use the label compliance.openshift.io/check-has-value that lists the variables a check has used.

Previously, while performing a scan, an unexpected termination occurred in one of the scanner containers of the pods. In this release, the Compliance Operator uses the latest OpenSCAP version 1.3.5 to avoid a crash.

Previously, using autoReplyRemediations to apply remediations triggered an update of the cluster nodes. This was disruptive if some of the remediations did not include all of the required input variables. Now, if a remediation is missing one or more required input variables, it is assigned a state of NeedsReview. If one or more remediations are in a NeedsReview state, the machine config pool remains paused, and the remediations are not applied until all of the required variables are set. This helps minimize disruption to the nodes.

The RBAC Role and Role Binding used for Prometheus metrics are changed to 'ClusterRole' and 'ClusterRoleBinding' to ensure that monitoring works without customization.

Previously, if an error occurred while parsing a profile, rules or variables objects were removed and deleted from the profile. Now, if an error occurs during parsing, the profileparser annotates the object with a temporary annotation that prevents the object from being deleted until after parsing completes. (BZ#1988259)

Previously, an error occurred if titles or descriptions were missing from a tailored profile. Because the XCCDF standard requires titles and descriptions for tailored profiles, titles and descriptions are now required to be set in TailoredProfile CRs.

Previously, when using tailored profiles, TailoredProfile variable values were allowed to be set using only a specific selection set. This restriction is now removed, and TailoredProfile variables can be set to any value.

Previously, the Compliance Operator was unable to parse Payment Card Industry Data Security Standard (PCI DSS) references. Now, the Operator can parse compliance content that ships with PCI DSS profiles.

Previously, the Compliance Operator was unable to execute rules for AU-5 control in the moderate profile. Now, permission is added to the Operator so that it can read Prometheusrules.monitoring.coreos.com objects and run the rules that cover AU-5 control in the moderate profile.