Kubernetes
uncordon_node
Uncordon nodes matching the given label name, so that pods can be scheduled on them again
Type | action |
Module | chaosk8s.node.actions |
Name | delete_nodes |
Return | None |
Usage
JSON
{
"name": "delete-nodes",
"type": "action",
"provider": {
"type": "python",
"module": "chaosk8s.node.actions",
"func": "delete_nodes"
}
}
YAML
name: delete-nodes
provider:
func: delete_nodes
module: chaosk8s.node.actions
type: python
type: action
Arguments
Name | Type | Default | Required | Title | Description |
---|---|---|---|---|---|
label_selector | string | Yes | Label Selector | Selectors to target the appropriate nodes | |
all | boolean | false | No | All Nodes | Delete all nodes matching the selector |
rand | boolean | false | No | Random Selection | Delete only a random selection matching the selector |
count | integer | 1 | No | Deletion Amount | Amount of nodes to delete |
grace_period_seconds | integer | No | Grace Period | Grace period for node termination | |
pod_label_selector | string | No | Per Pod Selection | Select nodes running the matching pods selection | |
pod_namespace | string | No | Pod Namespace | Pods selection namespace |
Select the appropriate nodes by label.
Nodes are not drained beforehand so we can see how the cluster behaves. Nodes cannot be restarted, they are deleted. Please be careful when using this action.
On certain cloud providers, you also need to delete the underneath VM instance as well afterward. This is the case with GCE for instance.
all
is set to True
, all nodes will be terminated.rand
is set to True
, one random node will be terminated.count
is set to a positive number, only a upto count
nodes (randomly picked) will be terminated. Otherwise, the first retrieved node will be terminated.Signature
def delete_nodes(label_selector: str = None,
all: bool = False,
rand: bool = False,
count: int = None,
grace_period_seconds: int = None,
secrets: Dict[str, Dict[str, str]] = None,
pod_label_selector: str = None,
pod_namespace: str = None):
pass