Kube-proxy: perf-enhancement: Reduce NAT table KUBE-SERVICES/NODEPORTS chain rules
The nat KUBE-SERVICES chain is called from OUTPUT and PREROUTING stages. In clusters with large number of services, the nat-KUBE-SERVICES chain is the largest chain with for eg: 33k rules. This patch aims to move the KubeMarkMasq rules from the kubeServicesChain into the respective KUBE-SVC-* chains. This way during each packet-rule matching we won't have to traverse the MASQ rules of all services which get accumulated in the KUBE-SERVICES and/or KUBE-NODEPORTS chains. Since the jump to KUBE-MARK-MASQ ultimately sets the 0x400 mark for nodeIP SNAT, it should not matter whether the jump is made from KUBE-SERVICES or KUBE-SVC-* chains. Specifically we change: 1) For ClusterIP svc, we move the KUBE-MARK-MASQ jump rule from KUBE-SERVICES chain into KUBE-SVC-* chain. 2) For ExternalIP svc, we move the KUBE-MARK-MASQ jump rule in the case of non-ServiceExternalTrafficPolicyTypeLocal from KUBE-SERVICES chain into KUBE-SVC-* chain. 3) For NodePorts svc, we move the KUBE-MARK-MASQ jump rule in case of non-ServiceExternalTrafficPolicyTypeLocal from KUBE-NODEPORTS chain to KUBE-SVC-* chain. 4) For load-balancer svc, we don't change anything since it is already svc specific due to creation of KUBE-FW-* chains per svc. This would cut the rules per svc in KUBE-SERVICES and KUBE-NODEPORTS in half.
This commit is contained in:
@@ -865,14 +865,14 @@ COMMIT
|
||||
-A KUBE-POSTROUTING -m comment --comment "kubernetes service traffic requiring SNAT" -j MASQUERADE
|
||||
-A KUBE-MARK-MASQ -j MARK --or-mark 0x4000
|
||||
-A KUBE-SERVICES -m comment --comment "ns1/svc1:p80 cluster IP" -m tcp -p tcp -d 10.20.30.41/32 --dport 80 -j KUBE-SVC-XPGD46QRK7WJZT7O
|
||||
-A KUBE-SERVICES -m comment --comment "ns1/svc1:p80 cluster IP" -m tcp -p tcp -d 10.20.30.41/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-XPGD46QRK7WJZT7O -m comment --comment "ns1/svc1:p80 cluster IP" -m tcp -p tcp -d 10.20.30.41/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-XPGD46QRK7WJZT7O -m comment --comment ns1/svc1:p80 -j KUBE-SEP-SXIVWICOYRO3J4NJ
|
||||
-A KUBE-SEP-SXIVWICOYRO3J4NJ -m comment --comment ns1/svc1:p80 -s 10.180.0.1/32 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SEP-SXIVWICOYRO3J4NJ -m comment --comment ns1/svc1:p80 -m tcp -p tcp -j DNAT --to-destination 10.180.0.1:80
|
||||
-A KUBE-SERVICES -m comment --comment "ns2/svc2:p80 cluster IP" -m tcp -p tcp -d 10.20.30.42/32 --dport 80 -j KUBE-SVC-GNZBNJ2PO5MGZ6GT
|
||||
-A KUBE-SERVICES -m comment --comment "ns2/svc2:p80 external IP" -m tcp -p tcp -d 1.2.3.4/32 --dport 80 -j KUBE-XLB-GNZBNJ2PO5MGZ6GT
|
||||
-A KUBE-SERVICES -m comment --comment "ns2/svc2:p80 loadbalancer IP" -m tcp -p tcp -d 1.2.3.4/32 --dport 80 -j KUBE-FW-GNZBNJ2PO5MGZ6GT
|
||||
-A KUBE-SERVICES -m comment --comment "ns2/svc2:p80 cluster IP" -m tcp -p tcp -d 10.20.30.42/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-GNZBNJ2PO5MGZ6GT -m comment --comment "ns2/svc2:p80 cluster IP" -m tcp -p tcp -d 10.20.30.42/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-GNZBNJ2PO5MGZ6GT -m comment --comment ns2/svc2:p80 -j KUBE-SEP-RS4RBKLTHTF2IUXJ
|
||||
-A KUBE-SEP-RS4RBKLTHTF2IUXJ -m comment --comment ns2/svc2:p80 -s 10.180.0.2/32 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SEP-RS4RBKLTHTF2IUXJ -m comment --comment ns2/svc2:p80 -m tcp -p tcp -j DNAT --to-destination 10.180.0.2:80
|
||||
@@ -885,16 +885,16 @@ COMMIT
|
||||
-A KUBE-XLB-GNZBNJ2PO5MGZ6GT -m comment --comment "route LOCAL traffic for ns2/svc2:p80 LB IP to service chain" -m addrtype --src-type LOCAL -j KUBE-SVC-GNZBNJ2PO5MGZ6GT
|
||||
-A KUBE-XLB-GNZBNJ2PO5MGZ6GT -m comment --comment "ns2/svc2:p80 has no local endpoints" -j KUBE-MARK-DROP
|
||||
-A KUBE-SERVICES -m comment --comment "ns3/svc3:p80 cluster IP" -m tcp -p tcp -d 10.20.30.43/32 --dport 80 -j KUBE-SVC-X27LE4BHSL4DOUIK
|
||||
-A KUBE-SERVICES -m comment --comment "ns3/svc3:p80 cluster IP" -m tcp -p tcp -d 10.20.30.43/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-X27LE4BHSL4DOUIK -m comment --comment "ns3/svc3:p80 cluster IP" -m tcp -p tcp -d 10.20.30.43/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-NODEPORTS -m comment --comment ns3/svc3:p80 -m tcp -p tcp --dport 3001 -j KUBE-SVC-X27LE4BHSL4DOUIK
|
||||
-A KUBE-NODEPORTS -m comment --comment ns3/svc3:p80 -m tcp -p tcp --dport 3001 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-X27LE4BHSL4DOUIK -m comment --comment ns3/svc3:p80 -m tcp -p tcp --dport 3001 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-X27LE4BHSL4DOUIK -m comment --comment ns3/svc3:p80 -j KUBE-SEP-OYPFS5VJICHGATKP
|
||||
-A KUBE-SEP-OYPFS5VJICHGATKP -m comment --comment ns3/svc3:p80 -s 10.180.0.3/32 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SEP-OYPFS5VJICHGATKP -m comment --comment ns3/svc3:p80 -m tcp -p tcp -j DNAT --to-destination 10.180.0.3:80
|
||||
-A KUBE-SERVICES -m comment --comment "ns4/svc4:p80 cluster IP" -m tcp -p tcp -d 10.20.30.44/32 --dport 80 -j KUBE-SVC-4SW47YFZTEDKD3PK
|
||||
-A KUBE-SERVICES -m comment --comment "ns4/svc4:p80 external IP" -m tcp -p tcp -d 50.60.70.81/32 --dport 80 -j KUBE-SVC-4SW47YFZTEDKD3PK
|
||||
-A KUBE-SERVICES -m comment --comment "ns4/svc4:p80 cluster IP" -m tcp -p tcp -d 10.20.30.44/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SERVICES -m comment --comment "ns4/svc4:p80 external IP" -m tcp -p tcp -d 50.60.70.81/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-4SW47YFZTEDKD3PK -m comment --comment "ns4/svc4:p80 cluster IP" -m tcp -p tcp -d 10.20.30.44/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-4SW47YFZTEDKD3PK -m comment --comment "ns4/svc4:p80 external IP" -m tcp -p tcp -d 50.60.70.81/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-4SW47YFZTEDKD3PK -m comment --comment ns4/svc4:p80 -m statistic --mode random --probability 0.5000000000 -j KUBE-SEP-UKSFD7AGPMPPLUHC
|
||||
-A KUBE-SVC-4SW47YFZTEDKD3PK -m comment --comment ns4/svc4:p80 -j KUBE-SEP-C6EBXVWJJZMIWKLZ
|
||||
-A KUBE-SEP-UKSFD7AGPMPPLUHC -m comment --comment ns4/svc4:p80 -s 10.180.0.4/32 -j KUBE-MARK-MASQ
|
||||
@@ -1007,6 +1007,10 @@ func TestClusterIPEndpointsJump(t *testing.T) {
|
||||
}
|
||||
|
||||
svcRules := ipt.GetRules(svcChain)
|
||||
if !hasJump(svcRules, string(KubeMarkMasqChain), svcIP, svcPort) {
|
||||
errorf(fmt.Sprintf("Failed to find jump from %v to KUBE-MARK-MASQ chain", svcChain), svcRules, t)
|
||||
}
|
||||
|
||||
if !hasJump(svcRules, epChain, "", 0) {
|
||||
errorf(fmt.Sprintf("Failed to jump to ep chain %v", epChain), svcRules, t)
|
||||
}
|
||||
@@ -1140,6 +1144,11 @@ func TestNodePort(t *testing.T) {
|
||||
if !hasJump(kubeNodePortRules, svcChain, "", svcNodePort) {
|
||||
errorf(fmt.Sprintf("Failed to find jump to svc chain %v", svcChain), kubeNodePortRules, t)
|
||||
}
|
||||
expectedNodePortNonLocalTrafficMasqueradeRule := `-A KUBE-SVC-XPGD46QRK7WJZT7O -m comment --comment ns1/svc1:p80 -m tcp -p tcp --dport 3001 -j KUBE-MARK-MASQ`
|
||||
svcRules := ipt.GetRules(svcChain)
|
||||
if !strings.Contains(fp.iptablesData.String(), expectedNodePortNonLocalTrafficMasqueradeRule) {
|
||||
errorf(fmt.Sprintf("Didn't find the masquerade rule for node port non-local traffic in svc chain %v", svcChain), svcRules, t)
|
||||
}
|
||||
}
|
||||
|
||||
func TestHealthCheckNodePort(t *testing.T) {
|
||||
@@ -1310,6 +1319,82 @@ func TestOnlyLocalExternalIPs(t *testing.T) {
|
||||
}
|
||||
}
|
||||
|
||||
// TestNonLocalExternalIPs tests if we add the masquerade rule into svcChain in order to
|
||||
// SNAT packets to external IPs if externalTrafficPolicy is cluster and the traffic is NOT Local.
|
||||
func TestNonLocalExternalIPs(t *testing.T) {
|
||||
ipt := iptablestest.NewFake()
|
||||
fp := NewFakeProxier(ipt, false)
|
||||
svcIP := "10.20.30.41"
|
||||
svcPort := 80
|
||||
svcExternalIPs := "50.60.70.81"
|
||||
svcPortName := proxy.ServicePortName{
|
||||
NamespacedName: makeNSN("ns1", "svc1"),
|
||||
Port: "p80",
|
||||
}
|
||||
|
||||
makeServiceMap(fp,
|
||||
makeTestService(svcPortName.Namespace, svcPortName.Name, func(svc *v1.Service) {
|
||||
svc.Spec.ClusterIP = svcIP
|
||||
svc.Spec.ExternalIPs = []string{svcExternalIPs}
|
||||
svc.Spec.Ports = []v1.ServicePort{{
|
||||
Name: svcPortName.Port,
|
||||
Port: int32(svcPort),
|
||||
Protocol: v1.ProtocolTCP,
|
||||
TargetPort: intstr.FromInt(svcPort),
|
||||
}}
|
||||
}),
|
||||
)
|
||||
makeEndpointsMap(fp)
|
||||
epIP1 := "10.180.0.1"
|
||||
epIP2 := "10.180.2.1"
|
||||
makeEndpointsMap(fp,
|
||||
makeTestEndpoints(svcPortName.Namespace, svcPortName.Name, func(ept *v1.Endpoints) {
|
||||
ept.Subsets = []v1.EndpointSubset{{
|
||||
Addresses: []v1.EndpointAddress{{
|
||||
IP: epIP1,
|
||||
NodeName: nil,
|
||||
}, {
|
||||
IP: epIP2,
|
||||
NodeName: utilpointer.StringPtr(testHostname),
|
||||
}},
|
||||
Ports: []v1.EndpointPort{{
|
||||
Name: svcPortName.Port,
|
||||
Port: int32(svcPort),
|
||||
Protocol: v1.ProtocolTCP,
|
||||
}},
|
||||
}}
|
||||
}),
|
||||
)
|
||||
|
||||
fp.syncProxyRules()
|
||||
proto := strings.ToLower(string(v1.ProtocolTCP))
|
||||
lbChain := string(serviceLBChainName(svcPortName.String(), proto))
|
||||
svcChain := string(servicePortChainName(svcPortName.String(), strings.ToLower(string(v1.ProtocolTCP))))
|
||||
|
||||
kubeSvcRules := ipt.GetRules(string(kubeServicesChain))
|
||||
if !hasJump(kubeSvcRules, svcChain, svcExternalIPs, svcPort) {
|
||||
errorf(fmt.Sprintf("Failed to find jump to svc chain %v", svcChain), kubeSvcRules, t)
|
||||
}
|
||||
|
||||
svcRules := ipt.GetRules(svcChain)
|
||||
if len(svcRules) != 4 {
|
||||
t.Errorf("expected svcChain %v to have 4 rules, got %v", svcChain, len(svcRules))
|
||||
}
|
||||
if !hasJump(svcRules, string(KubeMarkMasqChain), svcIP, svcPort) {
|
||||
errorf(fmt.Sprintf("Failed to find jump from %v to KUBE-MARK-MASQ chain", svcChain), svcRules, t)
|
||||
}
|
||||
expectedExternalIPNonLocalTrafficMasqueradeRule := `-A KUBE-SVC-XPGD46QRK7WJZT7O -m comment --comment "ns1/svc1:p80 external IP" -m tcp -p tcp -d 50.60.70.81/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ`
|
||||
|
||||
if !strings.Contains(fp.iptablesData.String(), expectedExternalIPNonLocalTrafficMasqueradeRule) {
|
||||
errorf(fmt.Sprintf("Didn't find the masquerade rule for external-ip non-local traffic in svc chain %v", svcChain), svcRules, t)
|
||||
}
|
||||
|
||||
lbRules := ipt.GetRules(lbChain)
|
||||
if len(lbRules) != 0 {
|
||||
t.Errorf("expected svclbChain %v to have 0 rules, got %v", lbChain, len(lbRules))
|
||||
}
|
||||
}
|
||||
|
||||
func TestNodePortReject(t *testing.T) {
|
||||
ipt := iptablestest.NewFake()
|
||||
fp := NewFakeProxier(ipt, false)
|
||||
@@ -2886,7 +2971,7 @@ COMMIT
|
||||
-A KUBE-POSTROUTING -j MARK --xor-mark 0x4000
|
||||
-A KUBE-POSTROUTING -m comment --comment "kubernetes service traffic requiring SNAT" -j MASQUERADE
|
||||
-A KUBE-MARK-MASQ -j MARK --or-mark 0x4000
|
||||
-A KUBE-SERVICES -m comment --comment "ns1/svc1 cluster IP" -m tcp -p tcp -d 172.20.1.1/32 --dport 0 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-AQI2S6QIMU7PVVRP -m comment --comment "ns1/svc1 cluster IP" -m tcp -p tcp -d 172.20.1.1/32 --dport 0 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SERVICES -m comment --comment "ns1/svc1 cluster IP" -m tcp -p tcp -d 172.20.1.1/32 --dport 0 -j KUBE-SVC-AQI2S6QIMU7PVVRP
|
||||
-A KUBE-SVC-AQI2S6QIMU7PVVRP -m comment --comment ns1/svc1 -m statistic --mode random --probability 0.3333333333 -j KUBE-SEP-3JOIVZTXZZRGORX4
|
||||
-A KUBE-SEP-3JOIVZTXZZRGORX4 -m comment --comment ns1/svc1 -s 10.0.1.1/32 -j KUBE-MARK-MASQ
|
||||
@@ -2986,7 +3071,7 @@ COMMIT
|
||||
-A KUBE-POSTROUTING -j MARK --xor-mark 0x4000
|
||||
-A KUBE-POSTROUTING -m comment --comment "kubernetes service traffic requiring SNAT" -j MASQUERADE
|
||||
-A KUBE-MARK-MASQ -j MARK --or-mark 0x4000
|
||||
-A KUBE-SERVICES -m comment --comment "ns1/svc1 cluster IP" -m tcp -p tcp -d 172.20.1.1/32 --dport 0 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-AQI2S6QIMU7PVVRP -m comment --comment "ns1/svc1 cluster IP" -m tcp -p tcp -d 172.20.1.1/32 --dport 0 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SERVICES -m comment --comment "ns1/svc1 cluster IP" -m tcp -p tcp -d 172.20.1.1/32 --dport 0 -j KUBE-SVC-AQI2S6QIMU7PVVRP
|
||||
-A KUBE-NODEPORTS -m comment --comment ns1/svc1 -m tcp -p tcp --dport 30010 -s 127.0.0.0/8 -j KUBE-MARK-MASQ
|
||||
-A KUBE-NODEPORTS -m comment --comment ns1/svc1 -m tcp -p tcp --dport 30010 -j KUBE-XLB-AQI2S6QIMU7PVVRP
|
||||
@@ -3309,7 +3394,7 @@ COMMIT
|
||||
-A KUBE-POSTROUTING -j MARK --xor-mark 0x4000
|
||||
-A KUBE-POSTROUTING -m comment --comment "kubernetes service traffic requiring SNAT" -j MASQUERADE
|
||||
-A KUBE-MARK-MASQ -j MARK --or-mark 0x4000
|
||||
-A KUBE-SERVICES -m comment --comment "ns1/svc1 cluster IP" -m tcp -p tcp -d 172.20.1.1/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-AQI2S6QIMU7PVVRP -m comment --comment "ns1/svc1 cluster IP" -m tcp -p tcp -d 172.20.1.1/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SERVICES -m comment --comment "ns1/svc1 cluster IP" -m tcp -p tcp -d 172.20.1.1/32 --dport 80 -j KUBE-SVC-AQI2S6QIMU7PVVRP
|
||||
-A KUBE-SVC-AQI2S6QIMU7PVVRP -m comment --comment ns1/svc1 -m statistic --mode random --probability 0.3333333333 -j KUBE-SEP-3JOIVZTXZZRGORX4
|
||||
-A KUBE-SEP-3JOIVZTXZZRGORX4 -m comment --comment ns1/svc1 -s 10.0.1.1/32 -j KUBE-MARK-MASQ
|
||||
@@ -3375,7 +3460,7 @@ COMMIT
|
||||
-A KUBE-POSTROUTING -j MARK --xor-mark 0x4000
|
||||
-A KUBE-POSTROUTING -m comment --comment "kubernetes service traffic requiring SNAT" -j MASQUERADE
|
||||
-A KUBE-MARK-MASQ -j MARK --or-mark 0x4000
|
||||
-A KUBE-SERVICES -m comment --comment "ns1/svc1 cluster IP" -m tcp -p tcp -d 172.20.1.1/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SVC-AQI2S6QIMU7PVVRP -m comment --comment "ns1/svc1 cluster IP" -m tcp -p tcp -d 172.20.1.1/32 --dport 80 ! -s 10.0.0.0/24 -j KUBE-MARK-MASQ
|
||||
-A KUBE-SERVICES -m comment --comment "ns1/svc1 cluster IP" -m tcp -p tcp -d 172.20.1.1/32 --dport 80 -j KUBE-SVC-AQI2S6QIMU7PVVRP
|
||||
-A KUBE-SVC-AQI2S6QIMU7PVVRP -m comment --comment ns1/svc1 -j KUBE-SEP-3JOIVZTXZZRGORX4
|
||||
-A KUBE-SEP-3JOIVZTXZZRGORX4 -m comment --comment ns1/svc1 -s 10.0.1.1/32 -j KUBE-MARK-MASQ
|
||||
|
Reference in New Issue
Block a user