mirror of
https://github.com/ubicloud/ubicloud.git
synced 2025-10-04 05:42:15 +08:00
08813d6fae
Prior to this commit, we would run the drain/delete command on the last node of the cluster but in a setup where we have 1 control plane node, by the time we are running the delete command, we have already run the kubeadm reset resulting in errors since there are no kubeconfigs there. With this commit, we make sure we are choosing a functional node and not a draining one.
90 lines
2.4 KiB
Ruby
90 lines
2.4 KiB
Ruby
# frozen_string_literal: true
|
|
|
|
class Prog::Kubernetes::KubernetesNodeNexus < Prog::Base
|
|
subject_is :kubernetes_node
|
|
|
|
def self.assemble(project_id, sshable_unix_user:, name:, location_id:, size:, storage_volumes:, boot_image:, private_subnet_id:, enable_ip4:, kubernetes_cluster_id:, kubernetes_nodepool_id: nil)
|
|
DB.transaction do
|
|
id = KubernetesNode.generate_uuid
|
|
vm = Prog::Vm::Nexus.assemble_with_sshable(project_id, sshable_unix_user:, name:, location_id:,
|
|
size:, storage_volumes:, boot_image:, private_subnet_id:, enable_ip4:).subject
|
|
KubernetesNode.create_with_id(id, vm_id: vm.id, kubernetes_cluster_id:, kubernetes_nodepool_id:)
|
|
Strand.create_with_id(id, prog: "Kubernetes::KubernetesNodeNexus", label: "start")
|
|
end
|
|
end
|
|
|
|
def before_run
|
|
when_destroy_set? do
|
|
if strand.label != "destroy"
|
|
hop_destroy
|
|
end
|
|
end
|
|
end
|
|
|
|
def cluster
|
|
kubernetes_node.kubernetes_cluster
|
|
end
|
|
|
|
def nodepool
|
|
kubernetes_node.kubernetes_nodepool
|
|
end
|
|
|
|
label def start
|
|
hop_wait
|
|
end
|
|
|
|
label def wait
|
|
when_retire_set? do
|
|
hop_retire
|
|
end
|
|
nap 6 * 60 * 60
|
|
end
|
|
|
|
label def retire
|
|
kubernetes_node.update(state: "draining")
|
|
hop_drain
|
|
end
|
|
|
|
label def drain
|
|
unit_name = "drain_node_#{kubernetes_node.name}"
|
|
sshable = cluster.sshable
|
|
case sshable.d_check(unit_name)
|
|
when "Succeeded"
|
|
hop_remove_node_from_cluster
|
|
when "NotStarted"
|
|
sshable.d_run(unit_name, "sudo", "kubectl", "--kubeconfig=/etc/kubernetes/admin.conf",
|
|
"drain", kubernetes_node.name, "--ignore-daemonsets", "--delete-emptydir-data")
|
|
nap 10
|
|
when "InProgress"
|
|
nap 10
|
|
when "Failed"
|
|
sshable.d_restart(unit_name)
|
|
nap 10
|
|
else
|
|
register_deadline("destroy", 0)
|
|
nap 3 * 60 * 60
|
|
end
|
|
end
|
|
|
|
label def remove_node_from_cluster
|
|
# kubeadm reset is necessary for etcd member removal, deleting the node itself
|
|
# won't remove the node from the etcd cluster, hurting the etcd cluster health
|
|
kubernetes_node.sshable.cmd("sudo kubeadm reset --force")
|
|
|
|
if nodepool
|
|
cluster.services_lb.detach_vm(kubernetes_node.vm)
|
|
else
|
|
cluster.api_server_lb.detach_vm(kubernetes_node.vm)
|
|
end
|
|
|
|
cluster.client.delete_node(kubernetes_node.name)
|
|
|
|
hop_destroy
|
|
end
|
|
|
|
label def destroy
|
|
kubernetes_node.vm.incr_destroy
|
|
kubernetes_node.destroy
|
|
pop "kubernetes node is deleted"
|
|
end
|
|
end
|