Files
ubicloud/spec/prog/kubernetes/kubernetes_cluster_nexus_spec.rb
Jeremy Evans 7544717b96 Have last exiting child schedule parent strand
Previously, when reaping child processes, if there were no remaining
reapable children, the parent strand would only nap 1, which puts
unnecessary load on respirate unless at least one child strand
exits in the next second.

Change this approach by having the exiting child strands, after they
release the lease, schedule their parent immediately if the parent
has no non-exited child strands.

When doing this, you need to be careful to make sure there are not
race conditions that would delay the scheduling of the parent.
There are two potential situations you need to handle:

1. Multiple children exiting at the same time
2. Parent currently running while child is exiting

By waiting until after the child strand leases are released, you
still have a race condition with 1, but the race condition is that
multiple child strands exiting concurrently could both reschedule
the parent strand.  However, that isn't a problem. You want to avoid
the case where neither child strand schedules the parent, which
rescheduling after releasing the lease should do.

To handle 2, inside reap use Model#lock! to lock the parent strand.
This will make exiting child strands block if they UPDATE the parent
strand with a new schedule, until the parent strand's transaction
commits.  However, it's possible that a child strand already UPDATED
the parent. To handle this situation, before calling lock!, store
the cached schedule value in a local variable.  lock! implicitly does
a reload, so compare the schedule value after reload.  If the schedule
has changed, likely a child scheduled the parent for immediate
execution, so nap 0 in that case.

Just in case there are unforeseen race conditions that are not handled,
only nap for 120 seconds if there are active children.  Worst case
scenario, this results in a 2 minute delay for running the parent.
However, this can potentially result in 120x less load from parent
strands polling children.
2025-06-28 03:30:43 +09:00

408 lines
20 KiB
Ruby

# frozen_string_literal: true
require_relative "../../model/spec_helper"
RSpec.describe Prog::Kubernetes::KubernetesClusterNexus do
subject(:nx) { described_class.new(st) }
let(:st) { Strand.new(id: "8148ebdf-66b8-8ed0-9c2f-8cfe93f5aa77") }
let(:customer_project) { Project.create(name: "default") }
let(:subnet) { PrivateSubnet.create(net6: "0::0", net4: "127.0.0.1", name: "x", location_id: Location::HETZNER_FSN1_ID, project_id: Config.kubernetes_service_project_id) }
let(:kubernetes_cluster) {
kc = KubernetesCluster.create(
name: "k8scluster",
version: "v1.32",
cp_node_count: 3,
private_subnet_id: subnet.id,
location_id: Location::HETZNER_FSN1_ID,
project_id: customer_project.id,
target_node_size: "standard-2"
)
KubernetesNodepool.create(name: "k8stest-np", node_count: 2, kubernetes_cluster_id: kc.id, target_node_size: "standard-2")
dns_zone = DnsZone.create(project_id: Project.first.id, name: "somezone", last_purged_at: Time.now)
apiserver_lb = LoadBalancer.create(private_subnet_id: subnet.id, name: "somelb", health_check_endpoint: "/foo", project_id: Config.kubernetes_service_project_id)
LoadBalancerPort.create(load_balancer_id: apiserver_lb.id, src_port: 123, dst_port: 456)
kc.add_cp_vm(create_vm)
kc.add_cp_vm(create_vm)
kc.update(api_server_lb_id: apiserver_lb.id)
services_lb = Prog::Vnet::LoadBalancerNexus.assemble(
subnet.id,
name: "somelb2",
algorithm: "hash_based",
# TODO: change the api to support LBs without ports
# The next two fields will be later modified by the sync_kubernetes_services label
# These are just set for passing the creation validations
src_port: 443,
dst_port: 6443,
health_check_endpoint: "/",
health_check_protocol: "tcp",
custom_hostname_dns_zone_id: dns_zone.id,
custom_hostname_prefix: "someprefix",
stack: LoadBalancer::Stack::IPV4
).subject
kc.update(services_lb_id: services_lb.id)
kc
}
before do
allow(Config).to receive(:kubernetes_service_project_id).and_return(Project.create(name: "UbicloudKubernetesService").id)
allow(nx).to receive(:kubernetes_cluster).and_return(kubernetes_cluster)
end
describe ".assemble" do
it "validates input" do
expect {
described_class.assemble(project_id: "88c8beda-0718-82d2-9948-7569acc26b80", name: "k8stest", version: "v1.32", location_id: Location::HETZNER_FSN1_ID, cp_node_count: 3, private_subnet_id: subnet.id)
}.to raise_error RuntimeError, "No existing project"
expect {
described_class.assemble(version: "v1.30", project_id: customer_project.id, name: "k8stest", location_id: Location::HETZNER_FSN1_ID, cp_node_count: 3, private_subnet_id: subnet.id)
}.to raise_error RuntimeError, "Invalid Kubernetes Version"
expect {
described_class.assemble(name: "Uppercase", version: "v1.32", project_id: customer_project.id, location_id: Location::HETZNER_FSN1_ID, cp_node_count: 3, private_subnet_id: subnet.id)
}.to raise_error Validation::ValidationFailed, "Validation failed for following fields: name"
expect {
described_class.assemble(name: "hyph_en", version: "v1.32", project_id: customer_project.id, location_id: Location::HETZNER_FSN1_ID, cp_node_count: 3, private_subnet_id: subnet.id)
}.to raise_error Validation::ValidationFailed, "Validation failed for following fields: name"
expect {
described_class.assemble(name: "onetoolongnameforatestkubernetesclustername", version: "v1.32", project_id: customer_project.id, location_id: Location::HETZNER_FSN1_ID, cp_node_count: 3, private_subnet_id: subnet.id)
}.to raise_error Validation::ValidationFailed, "Validation failed for following fields: name"
expect {
described_class.assemble(name: "somename", version: "v1.32", project_id: customer_project.id, location_id: Location::HETZNER_FSN1_ID, cp_node_count: 2, private_subnet_id: subnet.id)
}.to raise_error Validation::ValidationFailed, "Validation failed for following fields: control_plane_node_count"
p = Project.create(name: "another")
subnet.update(project_id: p.id)
expect {
described_class.assemble(name: "normalname", project_id: customer_project.id, location_id: Location::HETZNER_FSN1_ID, cp_node_count: 3, private_subnet_id: subnet.id)
}.to raise_error RuntimeError, "Given subnet is not available in the k8s project"
end
it "creates a kubernetes cluster" do
st = described_class.assemble(name: "k8stest", version: "v1.33", private_subnet_id: subnet.id, project_id: customer_project.id, location_id: Location::HETZNER_FSN1_ID, cp_node_count: 3, target_node_size: "standard-8", target_node_storage_size_gib: 100)
kc = st.subject
expect(kc.name).to eq "k8stest"
expect(kc.ubid).to start_with("kc")
expect(kc.version).to eq "v1.33"
expect(kc.location_id).to eq Location::HETZNER_FSN1_ID
expect(kc.cp_node_count).to eq 3
expect(kc.private_subnet.id).to eq subnet.id
expect(kc.project.id).to eq customer_project.id
expect(kc.strand.label).to eq "start"
expect(kc.target_node_size).to eq "standard-8"
expect(kc.target_node_storage_size_gib).to eq 100
end
it "has defaults for node size, storage size, version and subnet" do
st = described_class.assemble(name: "k8stest", project_id: customer_project.id, location_id: Location::HETZNER_FSN1_ID, cp_node_count: 3)
kc = st.subject
expect(kc.version).to eq "v1.32"
expect(kc.private_subnet.net4.to_s[-3..]).to eq "/18"
expect(kc.private_subnet.name).to eq kc.ubid.to_s + "-subnet"
expect(kc.target_node_size).to eq "standard-2"
expect(kc.target_node_storage_size_gib).to be_nil
end
end
describe "#before_run" do
it "hops to destroy" do
expect { nx.create_billing_records }.to hop("wait")
expect(nx).to receive(:when_destroy_set?).and_yield
expect(kubernetes_cluster.active_billing_records).not_to be_empty
expect(kubernetes_cluster.active_billing_records).to all(receive(:finalize))
expect { nx.before_run }.to hop("destroy")
end
it "does not hop to destroy if already in the destroy state" do
expect(nx).to receive(:when_destroy_set?).and_yield
expect(nx.strand).to receive(:label).and_return("destroy")
expect { nx.before_run }.not_to hop("destroy")
end
end
describe "#start" do
it "registers deadline and hops" do
expect(nx).to receive(:register_deadline)
expect { nx.start }.to hop("create_load_balancers")
end
end
describe "#create_load_balancers" do
it "creates api server and services load balancers with the right dns zone on prod and hops" do
kubernetes_cluster.update(api_server_lb_id: nil, services_lb_id: nil)
allow(Config).to receive(:kubernetes_service_hostname).and_return("k8s.ubicloud.com")
dns_zone = DnsZone.create(project_id: Project.first.id, name: "k8s.ubicloud.com", last_purged_at: Time.now)
expect { nx.create_load_balancers }.to hop("bootstrap_control_plane_vms")
expect(kubernetes_cluster.api_server_lb.name).to eq "#{kubernetes_cluster.ubid}-apiserver"
expect(kubernetes_cluster.api_server_lb.ports.first.src_port).to eq 443
expect(kubernetes_cluster.api_server_lb.ports.first.dst_port).to eq 6443
expect(kubernetes_cluster.api_server_lb.health_check_endpoint).to eq "/healthz"
expect(kubernetes_cluster.api_server_lb.health_check_protocol).to eq "tcp"
expect(kubernetes_cluster.api_server_lb.stack).to eq LoadBalancer::Stack::DUAL
expect(kubernetes_cluster.api_server_lb.private_subnet_id).to eq subnet.id
expect(kubernetes_cluster.api_server_lb.custom_hostname_dns_zone_id).to eq dns_zone.id
expect(kubernetes_cluster.api_server_lb.custom_hostname).to eq "k8scluster-apiserver-#{kubernetes_cluster.ubid[-5...]}.k8s.ubicloud.com"
expect(kubernetes_cluster.services_lb.name).to eq "#{kubernetes_cluster.ubid}-services"
expect(kubernetes_cluster.services_lb.stack).to eq LoadBalancer::Stack::IPV4
expect(kubernetes_cluster.services_lb.private_subnet_id).to eq subnet.id
expect(kubernetes_cluster.services_lb.custom_hostname_dns_zone_id).to eq dns_zone.id
expect(kubernetes_cluster.services_lb.custom_hostname).to eq "k8scluster-services-#{kubernetes_cluster.ubid[-5...]}.k8s.ubicloud.com"
end
it "creates load balancers with dns zone id on development for api server and services, then hops" do
expect { nx.create_load_balancers }.to hop("bootstrap_control_plane_vms")
expect(kubernetes_cluster.api_server_lb.name).to eq "#{kubernetes_cluster.ubid}-apiserver"
expect(kubernetes_cluster.api_server_lb.ports.first.src_port).to eq 443
expect(kubernetes_cluster.api_server_lb.ports.first.dst_port).to eq 6443
expect(kubernetes_cluster.api_server_lb.health_check_endpoint).to eq "/healthz"
expect(kubernetes_cluster.api_server_lb.health_check_protocol).to eq "tcp"
expect(kubernetes_cluster.api_server_lb.stack).to eq LoadBalancer::Stack::DUAL
expect(kubernetes_cluster.api_server_lb.private_subnet_id).to eq subnet.id
expect(kubernetes_cluster.api_server_lb.custom_hostname).to be_nil
expect(kubernetes_cluster.services_lb.name).to eq "#{kubernetes_cluster.ubid}-services"
expect(kubernetes_cluster.services_lb.private_subnet_id).to eq subnet.id
expect(kubernetes_cluster.services_lb.custom_hostname).to be_nil
end
end
describe "#bootstrap_control_plane_vms" do
it "waits until the load balancer endpoint is set" do
expect(kubernetes_cluster.api_server_lb).to receive(:hostname).and_return nil
expect { nx.bootstrap_control_plane_vms }.to nap(5)
end
it "incrs start_bootstrapping on KubernetesNodepool on 3 node control plane setup" do
expect(kubernetes_cluster).to receive(:cp_vms).and_return([create_vm, create_vm, create_vm]).twice
expect(kubernetes_cluster.nodepools.first).to receive(:incr_start_bootstrapping)
expect { nx.bootstrap_control_plane_vms }.to hop("wait_nodes")
end
it "incrs start_bootstrapping on KubernetesNodepool on 1 node control plane setup" do
kubernetes_cluster.update(cp_node_count: 1)
expect(kubernetes_cluster).to receive(:cp_vms).and_return([create_vm]).twice
expect(kubernetes_cluster.nodepools.first).to receive(:incr_start_bootstrapping)
expect { nx.bootstrap_control_plane_vms }.to hop("wait_nodes")
end
it "hops wait_nodes if the target number of CP vms is reached" do
expect(kubernetes_cluster.api_server_lb).to receive(:hostname).and_return "endpoint"
expect(kubernetes_cluster).to receive(:cp_vms).and_return([1, 2, 3]).twice
expect { nx.bootstrap_control_plane_vms }.to hop("wait_nodes")
end
it "buds ProvisionKubernetesNode prog to create VMs" do
expect(kubernetes_cluster).to receive(:endpoint).and_return "endpoint"
expect(nx).to receive(:bud).with(Prog::Kubernetes::ProvisionKubernetesNode, {"subject_id" => kubernetes_cluster.id})
expect { nx.bootstrap_control_plane_vms }.to hop("wait_control_plane_node")
end
end
describe "#wait_control_plane_node" do
it "hops back to bootstrap_control_plane_vms if there are no sub-programs running" do
st.update(prog: "Kubernetes::KubernetesClusterNexus", label: "wait_control_plane_node", stack: [{}])
expect { nx.wait_control_plane_node }.to hop("bootstrap_control_plane_vms")
end
it "donates if there are sub-programs running" do
st.update(prog: "Kubernetes::KubernetesClusterNexus", label: "wait_control_plane_node", stack: [{}])
Strand.create(parent_id: st.id, prog: "Kubernetes::ProvisionKubernetesNode", label: "start", stack: [{}], lease: Time.now + 10)
expect { nx.wait_control_plane_node }.to nap(120)
end
end
describe "#wait_nodes" do
it "naps until all nodepools are ready" do
expect(kubernetes_cluster.nodepools.first).to receive(:strand).and_return(instance_double(Strand, label: "not_wait"))
expect { nx.wait_nodes }.to nap(10)
end
it "hops to create_billing_records when all nodepools are ready" do
expect(kubernetes_cluster.nodepools.first).to receive(:strand).and_return(instance_double(Strand, label: "wait"))
expect { nx.wait_nodes }.to hop("create_billing_records")
end
end
describe "#create_billing_records" do
it "creates billing records for all cp vms and nodepools" do
kubernetes_cluster.nodepools.first.add_vm(create_vm)
expect { nx.create_billing_records }.to hop("wait")
expect(kubernetes_cluster.active_billing_records.length).to eq 4
expect(kubernetes_cluster.active_billing_records.map { it.billing_rate["resource_type"] }).to eq ["KubernetesControlPlaneVCpu", "KubernetesControlPlaneVCpu", "KubernetesWorkerVCpu", "KubernetesWorkerStorage"]
end
end
describe "#sync_kubernetes_services" do
it "calls the sync_kubernetes_services function" do
client = instance_double(Kubernetes::Client)
expect(nx).to receive(:decr_sync_kubernetes_services)
expect(kubernetes_cluster).to receive(:client).and_return(client)
expect(client).to receive(:sync_kubernetes_services)
expect { nx.sync_kubernetes_services }.to hop("wait")
end
end
describe "#wait" do
it "hops to the right sync_kubernetes_service when its semaphore is set" do
expect(nx).to receive(:when_sync_kubernetes_services_set?).and_yield
expect { nx.wait }.to hop("sync_kubernetes_services")
end
it "hops to upgrade when semaphore is set" do
expect(nx).to receive(:when_upgrade_set?).and_yield
expect { nx.wait }.to hop("upgrade")
end
it "naps until sync_kubernetes_service or upgrade is set" do
expect { nx.wait }.to nap(6 * 60 * 60)
end
end
describe "#upgrade" do
let(:first_vm) { create_vm }
let(:second_vm) { create_vm }
let(:client) { instance_double(Kubernetes::Client) }
before do
sshable0, sshable1 = instance_double(Sshable), instance_double(Sshable)
expect(kubernetes_cluster).to receive(:cp_vms).and_return([first_vm, second_vm])
allow(first_vm).to receive(:sshable).and_return(sshable0)
allow(second_vm).to receive(:sshable).and_return(sshable1)
allow(sshable0).to receive(:connect)
allow(sshable1).to receive(:connect)
expect(kubernetes_cluster).to receive(:client).and_return(client).at_least(:once)
end
it "selects a VM with minor version one less than the cluster's version" do
expect(kubernetes_cluster).to receive(:version).and_return("v1.32").twice
expect(client).to receive(:version).and_return("v1.32", "v1.31")
expect(nx).to receive(:bud).with(Prog::Kubernetes::UpgradeKubernetesNode, {"old_vm_id" => second_vm.id})
expect { nx.upgrade }.to hop("wait_upgrade")
end
it "hops to wait when all VMs are at the cluster's version" do
expect(kubernetes_cluster).to receive(:version).and_return("v1.32").twice
expect(client).to receive(:version).and_return("v1.32", "v1.32")
expect { nx.upgrade }.to hop("wait")
end
it "does not select a VM with minor version more than one less than the cluster's version" do
expect(kubernetes_cluster).to receive(:version).and_return("v1.32").twice
expect(client).to receive(:version).and_return("v1.30", "v1.32")
expect { nx.upgrade }.to hop("wait")
end
it "skips VMs with invalid version formats" do
expect(kubernetes_cluster).to receive(:version).and_return("v1.32").twice
expect(client).to receive(:version).and_return("invalid", "v1.32")
expect { nx.upgrade }.to hop("wait")
end
it "selects the first VM that is one minor version behind" do
expect(kubernetes_cluster).to receive(:version).and_return("v1.32")
expect(client).to receive(:version).and_return("v1.31")
expect(nx).to receive(:bud).with(Prog::Kubernetes::UpgradeKubernetesNode, {"old_vm_id" => first_vm.id})
expect { nx.upgrade }.to hop("wait_upgrade")
end
it "hops to wait if cluster version is invalid" do
expect(kubernetes_cluster).to receive(:version).and_return("invalid").twice
expect(client).to receive(:version).and_return("v1.31", "v1.31")
expect { nx.upgrade }.to hop("wait")
end
it "does not select a VM with a higher minor version than the cluster" do
expect(kubernetes_cluster).to receive(:version).and_return("v1.32").twice
expect(client).to receive(:version).and_return("v1.33", "v1.32")
expect { nx.upgrade }.to hop("wait")
end
end
describe "#wait_upgrade" do
it "hops back to upgrade if there are no sub-programs running" do
st.update(prog: "Kubernetes::KubernetesClusterNexus", label: "destroy", stack: [{}])
expect { nx.wait_upgrade }.to hop("upgrade")
end
it "donates if there are sub-programs running" do
st.update(prog: "Kubernetes::KubernetesClusterNexus", label: "destroy", stack: [{}])
Strand.create(parent_id: st.id, prog: "Kubernetes::ProvisionKubernetesNode", label: "start", stack: [{}], lease: Time.now + 10)
expect { nx.wait_upgrade }.to nap(120)
end
end
describe "#destroy" do
it "donates if there are sub-programs running (Provision...)" do
st.update(prog: "Kubernetes::KubernetesClusterNexus", label: "destroy", stack: [{}])
Strand.create(parent_id: st.id, prog: "Kubernetes::ProvisionKubernetesNode", label: "start", stack: [{}], lease: Time.now + 10)
expect { nx.destroy }.to nap(120)
end
it "triggers deletion of associated resources and naps until all nodepools are gone" do
st.update(prog: "Kubernetes::KubernetesClusterNexus", label: "destroy", stack: [{}])
expect(kubernetes_cluster.api_server_lb).to receive(:incr_destroy)
expect(kubernetes_cluster.services_lb).to receive(:incr_destroy)
expect(kubernetes_cluster.cp_vms).to all(receive(:incr_destroy))
expect(kubernetes_cluster.nodepools).to all(receive(:incr_destroy))
expect(kubernetes_cluster.private_subnet).to receive(:incr_destroy)
expect(kubernetes_cluster).not_to receive(:destroy)
expect { nx.destroy }.to nap(5)
end
it "completes destroy when nodepools are gone" do
st.update(prog: "Kubernetes::KubernetesClusterNexus", label: "destroy", stack: [{}])
kubernetes_cluster.nodepools.first.destroy
kubernetes_cluster.reload
expect(kubernetes_cluster.api_server_lb).to receive(:incr_destroy)
expect(kubernetes_cluster.services_lb).to receive(:incr_destroy)
expect(kubernetes_cluster.cp_vms).to all(receive(:incr_destroy))
expect(kubernetes_cluster.nodepools).to be_empty
expect { nx.destroy }.to exit({"msg" => "kubernetes cluster is deleted"})
end
it "deletes the sub-subdomain DNS record if the DNS zone exists" do
dns_zone = DnsZone.create_with_id(project_id: Project.first.id, name: "k8s.ubicloud.com", last_purged_at: Time.now)
kubernetes_cluster.services_lb.update(custom_hostname_dns_zone_id: dns_zone.id)
dns_zone.insert_record(record_name: "*.#{kubernetes_cluster.services_lb.hostname}.", type: "CNAME", ttl: 123, data: "whatever.")
expect(DnsRecord[name: "*.#{kubernetes_cluster.services_lb.hostname}.", tombstoned: false]).not_to be_nil
expect { nx.destroy }.to nap(5)
expect(DnsRecord[name: "*.#{kubernetes_cluster.services_lb.hostname}.", tombstoned: true]).not_to be_nil
end
it "completes the destroy process even if the load balancers do not exist" do
kubernetes_cluster.update(api_server_lb_id: nil, services_lb_id: nil)
kubernetes_cluster.nodepools.first.destroy
kubernetes_cluster.reload
expect { nx.destroy }.to exit({"msg" => "kubernetes cluster is deleted"})
end
end
end