utils

Documentation

Index

• func KubeletCommand(kOp KubeletOpt, c clientset.Interface, pod *v1.Pod)
• func PatchCSIDeployment(f *framework.Framework, o PatchCSIOptions, object interface{}) error
• func PodExec(pod *v1.Pod, bashExec string) (string, error)
• func PrivilegedTestPSPClusterRoleBinding(client clientset.Interface, namespace string, teardown bool, saNames []string)
• func RunInPodWithVolume(c clientset.Interface, ns, claimName, command string)
• func SIGDescribe(text string, body func()) bool
• func StartExternalProvisioner(c clientset.Interface, ns string, externalPluginName string) *v1.Pod
• func TestKubeletRestartsAndRestoresMount(c clientset.Interface, f *framework.Framework, clientPod *v1.Pod)
• func TestVolumeUnmountsFromDeletedPod(c clientset.Interface, f *framework.Framework, clientPod *v1.Pod)
• func TestVolumeUnmountsFromDeletedPodWithForceOption(c clientset.Interface, f *framework.Framework, clientPod *v1.Pod, ...)
• func TestVolumeUnmountsFromForceDeletedPod(c clientset.Interface, f *framework.Framework, clientPod *v1.Pod)
• func VerifyExecInPodFail(pod *v1.Pod, bashExec string, exitCode int)
• func VerifyExecInPodSucceed(pod *v1.Pod, bashExec string)
• type KubeletOpt
• type PatchCSIOptions

Functions

func KubeletCommand

func KubeletCommand(kOp KubeletOpt, c clientset.Interface, pod *v1.Pod)

KubeletCommand performs `start`, `restart`, or `stop` on the kubelet running on the node of the target pod and waits for the desired statues.. - First issues the command via `systemctl` - If `systemctl` returns stderr "command not found, issues the command via `service` - If `service` also returns stderr "command not found", the test is aborted. Allowed kubeletOps are `KStart`, `KStop`, and `KRestart`

func PatchCSIDeployment

func PatchCSIDeployment(f *framework.Framework, o PatchCSIOptions, object interface{}) error

PatchCSIDeployment modifies the CSI driver deployment: - replaces the provisioner name - forces pods onto a specific host

All of that is optional, see PatchCSIOptions. Just beware that not renaming the CSI driver deployment can be problematic:

• when multiple tests deploy the driver, they need to run sequentially
• might conflict with manual deployments

This function is written so that it works for CSI driver deployments that follow these conventions:

• driver and provisioner names are identical
• the driver binary accepts a --drivername parameter
• the provisioner binary accepts a --provisioner parameter
• the paths inside the container are either fixed and don't need to be patch (for example, --csi-address=/csi/csi.sock is okay) or are specified directly in a parameter (for example, --kubelet-registration-path=/var/lib/kubelet/plugins/csi-hostpath/csi.sock)

Driver deployments that are different will have to do the patching without this function, or skip patching entirely.

TODO (?): the storage.csi.image.version and storage.csi.image.registry settings are ignored. We could patch the image definitions or deprecate those options.

func PodExec

func PodExec(pod *v1.Pod, bashExec string) (string, error)

PodExec wraps RunKubectl to execute a bash cmd in target pod

func PrivilegedTestPSPClusterRoleBinding

func PrivilegedTestPSPClusterRoleBinding(client clientset.Interface,
	namespace string,
	teardown bool,
	saNames []string)

func RunInPodWithVolume

func RunInPodWithVolume(c clientset.Interface, ns, claimName, command string)

RunInPodWithVolume runs a command in a pod with given claim mounted to /mnt directory.

func SIGDescribe

func SIGDescribe(text string, body func()) bool

func StartExternalProvisioner

func StartExternalProvisioner(c clientset.Interface, ns string, externalPluginName string) *v1.Pod

func TestKubeletRestartsAndRestoresMount

func TestKubeletRestartsAndRestoresMount(c clientset.Interface, f *framework.Framework, clientPod *v1.Pod)

TestKubeletRestartsAndRestoresMount tests that a volume mounted to a pod remains mounted after a kubelet restarts

func TestVolumeUnmountsFromDeletedPod

func TestVolumeUnmountsFromDeletedPod(c clientset.Interface, f *framework.Framework, clientPod *v1.Pod)

TestVolumeUnmountsFromDeletedPod tests that a volume unmounts if the client pod was deleted while the kubelet was down.

func TestVolumeUnmountsFromDeletedPodWithForceOption

func TestVolumeUnmountsFromDeletedPodWithForceOption(c clientset.Interface, f *framework.Framework, clientPod *v1.Pod, forceDelete bool, checkSubpath bool)

TestVolumeUnmountsFromDeletedPod tests that a volume unmounts if the client pod was deleted while the kubelet was down. forceDelete is true indicating whether the pod is forcefully deleted.

func TestVolumeUnmountsFromForceDeletedPod

func TestVolumeUnmountsFromForceDeletedPod(c clientset.Interface, f *framework.Framework, clientPod *v1.Pod)

TestVolumeUnmountsFromFoceDeletedPod tests that a volume unmounts if the client pod was forcefully deleted while the kubelet was down.

func VerifyExecInPodFail

func VerifyExecInPodFail(pod *v1.Pod, bashExec string, exitCode int)

VerifyExecInPodFail verifies bash cmd in target pod fail with certain exit code

func VerifyExecInPodSucceed

func VerifyExecInPodSucceed(pod *v1.Pod, bashExec string)

VerifyExecInPodSucceed verifies bash cmd in target pod succeed

Types

type KubeletOpt

type KubeletOpt string

const (
	NodeStateTimeout            = 1 * time.Minute
	KStart           KubeletOpt = "start"
	KStop            KubeletOpt = "stop"
	KRestart         KubeletOpt = "restart"
)

type PatchCSIOptions

type PatchCSIOptions struct {
	// The original driver name.
	OldDriverName string
	// The driver name that replaces the original name.
	// Can be empty (not used at all) or equal to OldDriverName
	// (then it will be added were appropriate without renaming
	// in existing fields).
	NewDriverName string
	// The name of the container which has the CSI driver binary.
	// If non-empty, --drivername with the new name will be
	// appended to the argument list.
	DriverContainerName string
	// The name of the container which has the provisioner binary.
	// If non-empty, --provisioner with new name will be appended
	// to the argument list.
	ProvisionerContainerName string
	// If non-empty, all pods are forced to run on this node.
	NodeName string
}

PatchCSIOptions controls how PatchCSIDeployment patches the objects.