3장. PerformanceProfile [performance.openshift.io/v2]

additionalKernelArgs

array (string)

Additional kernel arguments.

cpu

object

CPU defines a set of CPU related parameters.

globallyDisableIrqLoadBalancing

boolean

GloballyDisableIrqLoadBalancing toggles whether IRQ load balancing will be disabled for the Isolated CPU set. When the option is set to "true" it disables IRQs load balancing for the Isolated CPU set. Setting the option to "false" allows the IRQs to be balanced across all CPUs, however the IRQs load balancing can be disabled per pod CPUs when using irq-load-balancing.crio.io/cpu-quota.crio.io annotations. Defaults to "false"

hardwareTuning

object

HardwareTuning defines a set of CPU frequencies for isolated and reserved cpus.

hugepages

object

HugePages defines a set of huge pages related parameters. It is possible to set huge pages with multiple size values at the same time. For example, hugepages can be set with 1G and 2M, both values will be set on the node by the Performance Profile Controller. It is important to notice that setting hugepages default size to 1G will remove all 2M related folders from the node and it will be impossible to configure 2M hugepages under the node.

kernelPageSize

string

KernelPageSize defines the kernel page size. 4k is the default, 64k is only supported on aarch64

machineConfigLabel

object (string)

MachineConfigLabel defines the label to add to the MachineConfigs the operator creates. It has to be used in the MachineConfigSelector of the MachineConfigPool which targets this performance profile. Defaults to "machineconfiguration.openshift.io/role=<same role as in NodeSelector label key>"

machineConfigPoolSelector

object (string)

MachineConfigPoolSelector defines the MachineConfigPool label to use in the MachineConfigPoolSelector of resources like KubeletConfigs created by the operator. Defaults to "machineconfiguration.openshift.io/role=<same role as in NodeSelector label key>"

net

object

Net defines a set of network related features

nodeSelector

object (string)

NodeSelector defines the Node label to use in the NodeSelectors of resources like Tuned created by the operator. It most likely should, but does not have to match the node label in the NodeSelector of the MachineConfigPool which targets this performance profile. In the case when machineConfigLabels or machineConfigPoolSelector are not set, we are expecting a certain NodeSelector format <domain>/<role>: "" in order to be able to calculate the default values for the former mentioned fields.

numa

object

NUMA defines options related to topology aware affinities

realTimeKernel

object

RealTimeKernel defines a set of real time kernel related parameters. RT kernel won’t be installed when not set.

workloadHints

object

WorkloadHints defines hints for different types of workloads. It will allow defining exact set of tuned and kernel arguments that should be applied on top of the node.

balanceIsolated

boolean

BalanceIsolated toggles whether or not the Isolated CPU set is eligible for load balancing work loads. When this option is set to "false", the Isolated CPU set will be static, meaning workloads have to explicitly assign each thread to a specific cpu in order to work across multiple CPUs. Setting this to "true" allows workloads to be balanced across CPUs. Setting this to "false" offers the most predictable performance for guaranteed workloads, but it offloads the complexity of cpu load balancing to the application. Defaults to "true"

isolated

string

Isolated defines a set of CPUs that will be used to give to application threads the most execution time possible, which means removing as many extraneous tasks off a CPU as possible. It is important to notice the CPU manager can choose any CPU to run the workload except the reserved CPUs. In order to guarantee that your workload will run on the isolated CPU: 1. The union of reserved CPUs and isolated CPUs should include all online CPUs 2. The isolated CPUs field should be the complementary to reserved CPUs field

offlined

string

Offline defines a set of CPUs that will be unused and set offline

reserved

string

Reserved defines a set of CPUs that will not be used for any container workloads initiated by kubelet.

shared

string

Shared defines a set of CPUs that will be shared among guaranteed workloads that needs additional cpus which are not exclusive, alongside the isolated, exclusive resources that are being used already by those workloads.

isolatedCpuFreq

integer

IsolatedCpuFreq defines a minimum frequency to be set across isolated cpus

reservedCpuFreq

integer

ReservedCpuFreq defines a maximum frequency to be set across reserved cpus

defaultHugepagesSize

string

DefaultHugePagesSize defines huge pages default size under kernel boot parameters.

pages

array

Pages defines huge pages that we want to allocate at boot time.

pages[]

object

HugePage defines the number of allocated huge pages of the specific size.

count

integer

Count defines amount of huge pages, maps to the 'hugepages' kernel boot parameter.

node

integer

Node defines the NUMA node where hugepages will be allocated, if not specified, pages will be allocated equally between NUMA nodes

size

string

Size defines huge page size, maps to the 'hugepagesz' kernel boot parameter.

devices

array

Devices contains a list of network device representations that will be set with a netqueue count equal to CPU.Reserved . If no devices are specified then the default is all devices.

devices[]

object

Device defines a way to represent a network device in several options: device name, vendor ID, model ID, PCI path and MAC address

userLevelNetworking

boolean

UserLevelNetworking when enabled - sets either all or specified network devices queue size to the amount of reserved CPUs. Defaults to "false".

deviceID

string

Network device ID (model) represnted as a 16 bit hexmadecimal number.

interfaceName

string

Network device name to be matched. It uses a syntax of shell-style wildcards which are either positive or negative.

vendorID

string

Network device vendor ID represnted as a 16 bit Hexmadecimal number.

topologyPolicy

string

Name of the policy applied when TopologyManager is enabled Operator defaults to "best-effort"

enabled

boolean

Enabled defines if the real time kernel packages should be installed. Defaults to "false"

highPowerConsumption

boolean

HighPowerConsumption defines if the node should be configured in high power consumption mode. The flag will affect the power consumption but will improve the CPUs latency. Defaults to false.

mixedCpus

boolean

MixedCpus enables the mixed-cpu-node-plugin on the node. Defaults to false.

perPodPowerManagement

boolean

PerPodPowerManagement defines if the node should be configured in per pod power management. PerPodPowerManagement and HighPowerConsumption hints can not be enabled together. Defaults to false.

realTime

boolean

RealTime defines if the node should be configured for the real time workload. Defaults to true.

conditions

array

Conditions represents the latest available observations of current state.

conditions[]

object

Condition represents the state of the operator’s reconciliation functionality.

runtimeClass

string

RuntimeClass contains the name of the RuntimeClass resource created by the operator.

tuned

string

Tuned points to the Tuned custom resource object that contains the tuning values generated by this operator.

lastHeartbeatTime

string

lastTransitionTime

string

message

string

reason

string

status

string

type

string

ConditionType is the state of the operator’s reconciliation functionality.