How to delete a node taint using Python's Kubernetes library - python-3.x

Question

Set the stain like this:
v3.patch_node('nodename',
{"spec": {"taints": [{"effect": "NoSchedule", "key": "test", "value": "1",'tolerationSeconds': '300'}]}}```
however.
How to remove tains?

Answer 1

This was pretty non-intuitive to me, but here's how I accomplished this.
def taint_node(context, node_name):
kube_client = setup_kube_client(context)
taint_patch = {"spec": {"taints": [{"effect": "NoSchedule", "key": "test", "value": "True"}]}}
return kube_client.patch_node(node_name, taint_patch)
def untaint_node(context, node_name):
kube_client = setup_kube_client(context)
remove_taint_patch = {"spec": {"taints": []}}
return kube_client.patch_node(node_name, remove_taint_patch)
That worked for me, but it removes ALL taints, which is maybe not what you want to do.
I tried the following:
def untaint_node(context, node_name):
kube_client = setup_kube_client(context)
remove_taint_patch = {"spec": {"taints": [{"effect": "NoSchedule-", "key": "test", "value": "True"}]}}
return kube_client.patch_node(node_name, remove_taint_patch)
but encountered server side validation preventing it (because the effect isn't in the collection of supported values):
kubernetes.client.exceptions.ApiException: (422)
Reason: Unprocessable Entity
HTTP response headers: HTTPHeaderDict({'Audit-Id': 'bfbad6e1-f37c-4090-898b-b2b9c5500425', 'Cache-Control': 'no-cache, private', 'Content-Type': 'application/json', 'X-Kubernetes-Pf-Flowschema-Uid': '7c028f53-f0a4-46bd-b400-68641158da78', 'X-Kubernetes-Pf-Prioritylevel-Uid': 'ef92e801-ce03-4abb-a607-20921bf82547', 'Date': 'Sat, 18 Sep 2021 02:45:37 GMT', 'Content-Length': '759'})
HTTP response body: {"kind":"Status","apiVersion":"v1","metadata":{},"status":"Failure","message":"Node \"aks-agentpool-33938206-vmss000000\" is invalid: metadata.taints[0].effect: Unsupported value: \"NoSchedule-\": supported values: \"NoSchedule\", \"PreferNoSchedule\", \"NoExecute\"","reason":"Invalid","details":{"name":"aks-agentpool-33938206-vmss000000","kind":"Node","causes":[{"reason":"FieldValueNotSupported","message":"Unsupported value: \"NoSchedule-\": supported values: \"NoSchedule\", \"PreferNoSchedule\", \"NoExecute\"","field":"metadata.taints[0].effect"},{"reason":"FieldValueNotSupported","message":"Unsupported value: \"NoSchedule-\": supported values: \"NoSchedule\", \"PreferNoSchedule\", \"NoExecute\"","field":"metadata.taints[0].effect"}]},"code":422}
Finally, if you need to remove a specific taint, you can always shell out to kubectl (though that's kinda cheating, huh?):
def untaint_node_with_cmd(context, node_name):
cmd_env = os.environ.copy()
child = subprocess.Popen(['kubectl', 'taint', 'nodes', node_name, 'test=True:NoSchedule-', '--context', context], env=cmd_env)
exit_code = child.wait()
return exit_code
Sadly, it doesn't look like this issue has gotten much love in the k8s python client repo. https://github.com/kubernetes-client/python/issues/161

Answer 2

One more better way to untainted a particular taint. By doing this way other taints will not get removed.only a particular taint will ve untainted.
def untaint_node(context, node_name, taint_key):
Kube_client = setup_kube_client(context)
node = Kube_client.list_nodes(field_selector={"metadata.name" : node_name}).items[0]
taints = node.spec.taints
filtered_taints = list(filter(lambda x: x.key != taint_key, taints))
body = {"spec": {"taints": filtered_taints}}
return kube_client.patch_node(node_name, body)

Answer 3

There's nothing special, standard update or patch call on the Node object.

Answer 4

Client libraries are used to interact with kubeapiserver. Therefore, kubeapiserver checks body of the request, no need to have custom removing taint in Python client library.
I think you can do it by calling
v3.patch_node('cn-shanghai.10.10.10.249',
{"spec": {"taints": [{"effect": "NoSchedule-", "key": "test", "value": "1","tolerationSeconds": "300"}]}}
An example can be found in python-client examples repository.
from kubernetes import client, config
def main():
config.load_kube_config()
api_instance = client.CoreV1Api()
body = {
"metadata": {
"labels": {
"foo": "bar",
"baz": None}
}
}
# Listing the cluster nodes
node_list = api_instance.list_node()
print("%s\t\t%s" % ("NAME", "LABELS"))
# Patching the node labels
for node in node_list.items:
api_response = api_instance.patch_node(node.metadata.name, body)
print("%s\t%s" % (node.metadata.name, node.metadata.labels))
if __name__ == '__main__':
main()
Reference: https://github.com/kubernetes-client/python/blob/c3f1a1c61efc608a4fe7f103ed103582c77bc30a/examples/node_labels.py