// NewProxier returns a new Proxier given an iptables Interface instance.
// Because of the iptables logic, it is assumed that there is only a single Proxier active on a machine.
// An error will be returned if iptables fails to update or acquire the initial lock.
// Once a proxier is created, it will keep iptables up to date in the background and
// will not terminate if a particular iptables call fails.
func NewProxier(ipt utiliptables.Interface, exec utilexec.Interface, syncPeriod time.Duration, masqueradeAll bool, masqueradeBit int) (*Proxier, error) {
// Set the route_localnet sysctl we need for
if err := utilsysctl.SetSysctl(sysctlRouteLocalnet, 1); err != nil {
return nil, fmt.Errorf("can't set sysctl %s: %v", sysctlRouteLocalnet, err)
}
// Load the module. It's OK if this fails (e.g. the module is not present)
// because we'll catch the error on the sysctl, which is what we actually
// care about.
exec.Command("modprobe", "br-netfilter").CombinedOutput()
if err := utilsysctl.SetSysctl(sysctlBridgeCallIptables, 1); err != nil {
glog.Warningf("can't set sysctl %s: %v", sysctlBridgeCallIptables, err)
}
// Generate the masquerade mark to use for SNAT rules.
if masqueradeBit < 0 || masqueradeBit > 31 {
return nil, fmt.Errorf("invalid iptables-masquerade-bit %v not in [0, 31]", masqueradeBit)
}
masqueradeValue := 1 << uint(masqueradeBit)
masqueradeMark := fmt.Sprintf("%#08x/%#08x", masqueradeValue, masqueradeValue)
return &Proxier{
serviceMap: make(map[proxy.ServicePortName]*serviceInfo),
endpointsMap: make(map[proxy.ServicePortName][]string),
portsMap: make(map[localPort]closeable),
syncPeriod: syncPeriod,
iptables: ipt,
masqueradeAll: masqueradeAll,
masqueradeMark: masqueradeMark,
}, nil
}
func StartNode(nodeConfig configapi.NodeConfig) error {
config, err := kubernetes.BuildKubernetesNodeConfig(nodeConfig)
if err != nil {
return err
}
glog.Infof("Starting node %s (%s)", config.KubeletServer.HostnameOverride, version.Get().String())
// preconditions
config.EnsureVolumeDir()
config.EnsureDocker(docker.NewHelper())
// async starts
config.RunKubelet()
config.RunSDN()
config.RunProxy()
// HACK: RunProxy resets bridge-nf-call-iptables from what openshift-sdn requires
if config.SDNPlugin != nil {
sdnPluginName := nodeConfig.NetworkConfig.NetworkPluginName
if sdnPluginName == osdn.SingleTenantPluginName() || sdnPluginName == osdn.MultiTenantPluginName() {
if err := sysctl.SetSysctl("net/bridge/bridge-nf-call-iptables", 0); err != nil {
glog.Warningf("Could not set net.bridge.bridge-nf-call-iptables sysctl: %s", err)
}
}
}
return nil
}
// setupKernelTunables validates kernel tunable flags are set as expected
// depending upon the specified option, it will either warn, error, or modify the kernel tunable flags
func setupKernelTunables(option KernelTunableBehavior) error {
desiredState := map[string]int{
utilsysctl.VmOvercommitMemory: utilsysctl.VmOvercommitMemoryAlways,
utilsysctl.VmPanicOnOOM: utilsysctl.VmPanicOnOOMInvokeOOMKiller,
utilsysctl.KernelPanic: utilsysctl.KernelPanicRebootTimeout,
utilsysctl.KernelPanicOnOops: utilsysctl.KernelPanicOnOopsAlways,
}
errList := []error{}
for flag, expectedValue := range desiredState {
val, err := utilsysctl.GetSysctl(flag)
if err != nil {
errList = append(errList, err)
continue
}
if val == expectedValue {
continue
}
switch option {
case KernelTunableError:
errList = append(errList, fmt.Errorf("Invalid kernel flag: %v, expected value: %v, actual value: %v", flag, expectedValue, val))
case KernelTunableWarn:
glog.V(2).Infof("Invalid kernel flag: %v, expected value: %v, actual value: %v", flag, expectedValue, val)
case KernelTunableModify:
glog.V(2).Infof("Updating kernel flag: %v, expected value: %v, actual value: %v", flag, expectedValue, val)
err = utilsysctl.SetSysctl(flag, expectedValue)
if err != nil {
errList = append(errList, err)
}
}
}
return utilerrors.NewAggregate(errList)
}
func (plugin *kubenetNetworkPlugin) Init(host network.Host) error {
plugin.host = host
plugin.cniConfig = &libcni.CNIConfig{
Path: []string{DefaultCNIDir},
}
if link, err := findMinMTU(); err == nil {
plugin.MTU = link.MTU
glog.V(5).Infof("Using interface %s MTU %d as bridge MTU", link.Name, link.MTU)
} else {
glog.Warningf("Failed to find default bridge MTU: %v", err)
}
// Since this plugin uses a Linux bridge, set bridge-nf-call-iptables=1
// is necessary to ensure kube-proxy functions correctly.
//
// This will return an error on older kernel version (< 3.18) as the module
// was built-in, we simply ignore the error here. A better thing to do is
// to check the kernel version in the future.
utilexec.New().Command("modprobe", "br-netfilter").CombinedOutput()
if err := utilsysctl.SetSysctl(sysctlBridgeCallIptables, 1); err != nil {
glog.Warningf("can't set sysctl %s: %v", sysctlBridgeCallIptables, err)
}
return nil
}
// NewProxier returns a new Proxier given an iptables Interface instance.
// Because of the iptables logic, it is assumed that there is only a single Proxier active on a machine.
// An error will be returned if iptables fails to update or acquire the initial lock.
// Once a proxier is created, it will keep iptables up to date in the background and
// will not terminate if a particular iptables call fails.
func NewProxier(ipt utiliptables.Interface, exec utilexec.Interface, syncPeriod time.Duration, masqueradeAll bool, masqueradeBit int, clusterCIDR string) (*Proxier, error) {
// Set the route_localnet sysctl we need for
if err := utilsysctl.SetSysctl(sysctlRouteLocalnet, 1); err != nil {
return nil, fmt.Errorf("can't set sysctl %s: %v", sysctlRouteLocalnet, err)
}
// Proxy needs br_netfilter and bridge-nf-call-iptables=1 when containers
// are connected to a Linux bridge (but not SDN bridges). Until most
// plugins handle this, log when config is missing
if val, err := utilsysctl.GetSysctl(sysctlBridgeCallIptables); err == nil && val != 1 {
glog.Infof("missing br-netfilter module or unset sysctl br-nf-call-iptables; proxy may not work as intended")
}
// Generate the masquerade mark to use for SNAT rules.
if masqueradeBit < 0 || masqueradeBit > 31 {
return nil, fmt.Errorf("invalid iptables-masquerade-bit %v not in [0, 31]", masqueradeBit)
}
masqueradeValue := 1 << uint(masqueradeBit)
masqueradeMark := fmt.Sprintf("%#08x/%#08x", masqueradeValue, masqueradeValue)
return &Proxier{
serviceMap: make(map[proxy.ServicePortName]*serviceInfo),
endpointsMap: make(map[proxy.ServicePortName][]string),
portsMap: make(map[localPort]closeable),
syncPeriod: syncPeriod,
iptables: ipt,
masqueradeAll: masqueradeAll,
masqueradeMark: masqueradeMark,
exec: exec,
clusterCIDR: clusterCIDR,
}, nil
}
// ResetSysctlFromProxy resets the bridge-nf-call-iptables systctl that the Kube proxy sets, which
// is required for normal Docker containers to talk to the SDN plugin on the local system.
// Resolution is https://github.com/kubernetes/kubernetes/pull/20647
func (c *NodeConfig) ResetSysctlFromProxy() {
if c.SDNPlugin == nil {
return
}
if err := sysctl.SetSysctl("net/bridge/bridge-nf-call-iptables", 0); err != nil {
glog.Warningf("Could not set net.bridge.bridge-nf-call-iptables sysctl: %s", err)
}
}
func (realConntracker) SetMax(max int) error {
glog.Infof("Setting nf_conntrack_max to %d", max)
if err := sysctl.SetSysctl("net/netfilter/nf_conntrack_max", max); err != nil {
return err
}
// TODO: generify this and sysctl to a new sysfs.WriteInt()
glog.Infof("Setting conntrack hashsize to %d", max/4)
return ioutil.WriteFile("/sys/module/nf_conntrack/parameters/hashsize", []byte(strconv.Itoa(max/4)), 0640)
}
// changeSysctl changes the required network setting in /proc to get
// keepalived working in the local system.
func changeSysctl() error {
for k, v := range sysctlAdjustments {
if err := sysctl.SetSysctl(k, v); err != nil {
return err
}
}
return nil
}
// NewProxier returns a new Proxier given an iptables Interface instance.
// Because of the iptables logic, it is assumed that there is only a single Proxier active on a machine.
// An error will be returned if iptables fails to update or acquire the initial lock.
// Once a proxier is created, it will keep iptables up to date in the background and
// will not terminate if a particular iptables call fails.
func NewProxier(ipt utiliptables.Interface, exec utilexec.Interface, syncPeriod time.Duration, masqueradeAll bool) (*Proxier, error) {
// Set the route_localnet sysctl we need for
if err := utilsysctl.SetSysctl(sysctlRouteLocalnet, 1); err != nil {
return nil, fmt.Errorf("can't set sysctl %s: %v", sysctlRouteLocalnet, err)
}
// Load the module. It's OK if this fails (e.g. the module is not present)
// because we'll catch the error on the sysctl, which is what we actually
// care about.
exec.Command("modprobe", "br-netfilter").CombinedOutput()
if err := utilsysctl.SetSysctl(sysctlBridgeCallIptables, 1); err != nil {
glog.Warningf("can't set sysctl %s: %v", sysctlBridgeCallIptables, err)
}
return &Proxier{
serviceMap: make(map[proxy.ServicePortName]*serviceInfo),
endpointsMap: make(map[proxy.ServicePortName][]string),
portsMap: make(map[localPort]closeable),
syncPeriod: syncPeriod,
iptables: ipt,
masqueradeAll: masqueradeAll,
}, nil
}
// disableKernelMemoryOvercommitHandling tells the kernel to perform no memory over-commit handling.
// Under this setting, the potential for memory overload is increased, but so is performance for
// memory-intensive tasks
// sets /proc/sys/vm/overcommit_memory to 1
func disableKernelMemoryOvercommitHandling() error {
val, err := utilsysctl.GetSysctl(sysctlVmOvercommitMemory)
if err != nil {
return err
}
if val == 1 {
return nil
}
glog.V(2).Infof("Updating kernel memory overcommit flag from %v to %v", val, 1)
err = utilsysctl.SetSysctl(sysctlVmOvercommitMemory, 1)
if err != nil {
return err
}
return nil
}
func (plugin *NoopNetworkPlugin) Init(host Host) error {
// Set bridge-nf-call-iptables=1 to maintain compatibility with older
// kubernetes versions to ensure the iptables-based kube proxy functions
// correctly. Other plugins are responsible for setting this correctly
// depending on whether or not they connect containers to Linux bridges
// or use some other mechanism (ie, SDN vswitch).
// Ensure the netfilter module is loaded on kernel >= 3.18; previously
// it was built-in.
utilexec.New().Command("modprobe", "br-netfilter").CombinedOutput()
if err := utilsysctl.SetSysctl(sysctlBridgeCallIptables, 1); err != nil {
glog.Warningf("can't set sysctl %s: %v", sysctlBridgeCallIptables, err)
}
return nil
}
func (plugin *kubenetNetworkPlugin) Init(host network.Host, hairpinMode componentconfig.HairpinMode, nonMasqueradeCIDR string) error {
plugin.host = host
plugin.hairpinMode = hairpinMode
plugin.nonMasqueradeCIDR = nonMasqueradeCIDR
plugin.cniConfig = &libcni.CNIConfig{
Path: []string{DefaultCNIDir, plugin.vendorDir},
}
if link, err := findMinMTU(); err == nil {
plugin.MTU = link.MTU
glog.V(5).Infof("Using interface %s MTU %d as bridge MTU", link.Name, link.MTU)
} else {
glog.Warningf("Failed to find default bridge MTU: %v", err)
}
// Since this plugin uses a Linux bridge, set bridge-nf-call-iptables=1
// is necessary to ensure kube-proxy functions correctly.
//
// This will return an error on older kernel version (< 3.18) as the module
// was built-in, we simply ignore the error here. A better thing to do is
// to check the kernel version in the future.
plugin.execer.Command("modprobe", "br-netfilter").CombinedOutput()
err := utilsysctl.SetSysctl(sysctlBridgeCallIptables, 1)
if err != nil {
glog.Warningf("can't set sysctl %s: %v", sysctlBridgeCallIptables, err)
}
plugin.loConfig, err = libcni.ConfFromBytes([]byte(`{
"cniVersion": "0.1.0",
"name": "kubenet-loopback",
"type": "loopback"
}`))
if err != nil {
return fmt.Errorf("Failed to generate loopback config: %v", err)
}
plugin.nsenterPath, err = plugin.execer.LookPath("nsenter")
if err != nil {
return fmt.Errorf("Failed to find nsenter binary: %v", err)
}
// Need to SNAT outbound traffic from cluster
if err = plugin.ensureMasqRule(); err != nil {
return err
}
return nil
}
func (realConntracker) SetMax(max int) error {
glog.Infof("Setting nf_conntrack_max to %d", max)
if err := sysctl.SetSysctl("net/netfilter/nf_conntrack_max", max); err != nil {
return err
}
// sysfs is expected to be mounted as 'rw'. However, it may be unexpectedly mounted as
// 'ro' by docker because of a known docker issue (https://github.com/docker/docker/issues/24000).
// Setting conntrack will fail when sysfs is readonly. When that happens, we don't set conntrack
// hashsize and return a special error readOnlySysFSError here. The caller should deal with
// readOnlySysFSError differently.
writable, err := isSysFSWritable()
if err != nil {
return err
}
if !writable {
return readOnlySysFSError
}
// TODO: generify this and sysctl to a new sysfs.WriteInt()
glog.Infof("Setting conntrack hashsize to %d", max/4)
return ioutil.WriteFile("/sys/module/nf_conntrack/parameters/hashsize", []byte(strconv.Itoa(max/4)), 0640)
}
func (plugin *OsdnNode) SetupSDN() (bool, error) {
clusterNetworkCIDR := plugin.networkInfo.ClusterNetwork.String()
serviceNetworkCIDR := plugin.networkInfo.ServiceNetwork.String()
localSubnetCIDR := plugin.localSubnetCIDR
_, ipnet, err := net.ParseCIDR(localSubnetCIDR)
localSubnetMaskLength, _ := ipnet.Mask.Size()
localSubnetGateway := netutils.GenerateDefaultGateway(ipnet).String()
glog.V(5).Infof("[SDN setup] node pod subnet %s gateway %s", ipnet.String(), localSubnetGateway)
exec := kexec.New()
if plugin.clearLbr0IptablesRule {
// Delete docker's left-over lbr0 rule; cannot do this from
// NewNodePlugin (where docker is cleaned up) because we need
// localSubnetCIDR which is only valid after plugin start
ipt := iptables.New(exec, utildbus.New(), iptables.ProtocolIpv4)
ipt.DeleteRule(iptables.TableNAT, iptables.ChainPostrouting, "-s", localSubnetCIDR, "!", "-o", "lbr0", "-j", "MASQUERADE")
}
gwCIDR := fmt.Sprintf("%s/%d", localSubnetGateway, localSubnetMaskLength)
if plugin.alreadySetUp(gwCIDR, clusterNetworkCIDR) {
glog.V(5).Infof("[SDN setup] no SDN setup required")
return false, nil
}
glog.V(5).Infof("[SDN setup] full SDN setup required")
if err := os.MkdirAll("/run/openshift-sdn", 0700); err != nil {
return false, err
}
config := fmt.Sprintf("export OPENSHIFT_CLUSTER_SUBNET=%s", clusterNetworkCIDR)
err = ioutil.WriteFile("/run/openshift-sdn/config.env", []byte(config), 0644)
if err != nil {
return false, err
}
err = plugin.ovs.AddBridge("fail-mode=secure", "protocols=OpenFlow13")
if err != nil {
return false, err
}
_ = plugin.ovs.DeletePort(VXLAN)
_, err = plugin.ovs.AddPort(VXLAN, 1, "type=vxlan", `options:remote_ip="flow"`, `options:key="flow"`)
if err != nil {
return false, err
}
_ = plugin.ovs.DeletePort(TUN)
_, err = plugin.ovs.AddPort(TUN, 2, "type=internal")
if err != nil {
return false, err
}
otx := plugin.ovs.NewTransaction()
// Table 0: initial dispatch based on in_port
// vxlan0
otx.AddFlow("table=0, priority=200, in_port=1, arp, nw_src=%s, nw_dst=%s, actions=move:NXM_NX_TUN_ID[0..31]->NXM_NX_REG0[],goto_table:1", clusterNetworkCIDR, localSubnetCIDR)
otx.AddFlow("table=0, priority=200, in_port=1, ip, nw_src=%s, nw_dst=%s, actions=move:NXM_NX_TUN_ID[0..31]->NXM_NX_REG0[],goto_table:1", clusterNetworkCIDR, localSubnetCIDR)
otx.AddFlow("table=0, priority=150, in_port=1, actions=drop")
// tun0
otx.AddFlow("table=0, priority=200, in_port=2, arp, nw_src=%s, nw_dst=%s, actions=goto_table:5", localSubnetGateway, clusterNetworkCIDR)
otx.AddFlow("table=0, priority=200, in_port=2, ip, actions=goto_table:5")
otx.AddFlow("table=0, priority=150, in_port=2, actions=drop")
// else, from a container
otx.AddFlow("table=0, priority=100, arp, actions=goto_table:2")
otx.AddFlow("table=0, priority=100, ip, actions=goto_table:2")
otx.AddFlow("table=0, priority=0, actions=drop")
// Table 1: VXLAN ingress filtering; filled in by AddHostSubnetRules()
// eg, "table=1, priority=100, tun_src=${remote_node_ip}, actions=goto_table:5"
otx.AddFlow("table=1, priority=0, actions=drop")
// Table 2: from OpenShift container; validate IP/MAC, assign tenant-id; filled in by openshift-sdn-ovs
// eg, "table=2, priority=100, in_port=${ovs_port}, arp, nw_src=${ipaddr}, arp_sha=${macaddr}, actions=load:${tenant_id}->NXM_NX_REG0[], goto_table:5"
// "table=2, priority=100, in_port=${ovs_port}, ip, nw_src=${ipaddr}, actions=load:${tenant_id}->NXM_NX_REG0[], goto_table:3"
// (${tenant_id} is always 0 for single-tenant)
otx.AddFlow("table=2, priority=0, actions=drop")
// Table 3: from OpenShift container; service vs non-service
otx.AddFlow("table=3, priority=100, ip, nw_dst=%s, actions=goto_table:4", serviceNetworkCIDR)
otx.AddFlow("table=3, priority=0, actions=goto_table:5")
// Table 4: from OpenShift container; service dispatch; filled in by AddServiceRules()
otx.AddFlow("table=4, priority=200, reg0=0, actions=output:2")
// eg, "table=4, priority=100, reg0=${tenant_id}, ${service_proto}, nw_dst=${service_ip}, tp_dst=${service_port}, actions=output:2"
otx.AddFlow("table=4, priority=0, actions=drop")
// Table 5: general routing
otx.AddFlow("table=5, priority=300, arp, nw_dst=%s, actions=output:2", localSubnetGateway)
otx.AddFlow("table=5, priority=300, ip, nw_dst=%s, actions=output:2", localSubnetGateway)
otx.AddFlow("table=5, priority=200, arp, nw_dst=%s, actions=goto_table:6", localSubnetCIDR)
otx.AddFlow("table=5, priority=200, ip, nw_dst=%s, actions=goto_table:7", localSubnetCIDR)
otx.AddFlow("table=5, priority=100, arp, nw_dst=%s, actions=goto_table:8", clusterNetworkCIDR)
otx.AddFlow("table=5, priority=100, ip, nw_dst=%s, actions=goto_table:8", clusterNetworkCIDR)
otx.AddFlow("table=5, priority=0, ip, actions=goto_table:9")
otx.AddFlow("table=5, priority=0, arp, actions=drop")
// Table 6: ARP to container, filled in by openshift-sdn-ovs
// eg, "table=6, priority=100, arp, nw_dst=${container_ip}, actions=output:${ovs_port}"
otx.AddFlow("table=6, priority=0, actions=drop")
// Table 7: IP to container; filled in by openshift-sdn-ovs
// eg, "table=7, priority=100, reg0=0, ip, nw_dst=${ipaddr}, actions=output:${ovs_port}"
// eg, "table=7, priority=100, reg0=${tenant_id}, ip, nw_dst=${ipaddr}, actions=output:${ovs_port}"
otx.AddFlow("table=7, priority=0, actions=drop")
// Table 8: to remote container; filled in by AddHostSubnetRules()
// eg, "table=8, priority=100, arp, nw_dst=${remote_subnet_cidr}, actions=move:NXM_NX_REG0[]->NXM_NX_TUN_ID[0..31], set_field:${remote_node_ip}->tun_dst,output:1"
// eg, "table=8, priority=100, ip, nw_dst=${remote_subnet_cidr}, actions=move:NXM_NX_REG0[]->NXM_NX_TUN_ID[0..31], set_field:${remote_node_ip}->tun_dst,output:1"
otx.AddFlow("table=8, priority=0, actions=drop")
// Table 9: egress network policy dispatch; edited by updateEgressNetworkPolicyRules()
// eg, "table=9, reg0=${tenant_id}, priority=2, ip, nw_dst=${external_cidr}, actions=drop
otx.AddFlow("table=9, priority=0, actions=output:2")
err = otx.EndTransaction()
if err != nil {
return false, err
}
itx := ipcmd.NewTransaction(exec, TUN)
itx.AddAddress(gwCIDR)
defer deleteLocalSubnetRoute(TUN, localSubnetCIDR)
itx.SetLink("mtu", fmt.Sprint(plugin.mtu))
itx.SetLink("up")
itx.AddRoute(clusterNetworkCIDR, "proto", "kernel", "scope", "link")
itx.AddRoute(serviceNetworkCIDR)
err = itx.EndTransaction()
if err != nil {
return false, err
}
sysctl := sysctl.New()
// Enable IP forwarding for ipv4 packets
err = sysctl.SetSysctl("net/ipv4/ip_forward", 1)
if err != nil {
return false, fmt.Errorf("Could not enable IPv4 forwarding: %s", err)
}
err = sysctl.SetSysctl(fmt.Sprintf("net/ipv4/conf/%s/forwarding", TUN), 1)
if err != nil {
return false, fmt.Errorf("Could not enable IPv4 forwarding on %s: %s", TUN, err)
}
// Table 253: rule version; note action is hex bytes separated by '.'
otx = plugin.ovs.NewTransaction()
pluginVersion := getPluginVersion(plugin.multitenant)
otx.AddFlow("%s, %s%s.%s", VERSION_TABLE, VERSION_ACTION, pluginVersion[0], pluginVersion[1])
err = otx.EndTransaction()
if err != nil {
return false, err
}
return true, nil
}
func (c *FlowController) Setup(localSubnetCIDR, clusterNetworkCIDR, servicesNetworkCIDR string, mtu uint) (bool, error) {
_, ipnet, err := net.ParseCIDR(localSubnetCIDR)
localSubnetMaskLength, _ := ipnet.Mask.Size()
localSubnetGateway := netutils.GenerateDefaultGateway(ipnet).String()
glog.V(5).Infof("[SDN setup] node pod subnet %s gateway %s", ipnet.String(), localSubnetGateway)
gwCIDR := fmt.Sprintf("%s/%d", localSubnetGateway, localSubnetMaskLength)
if alreadySetUp(c.multitenant, gwCIDR) {
glog.V(5).Infof("[SDN setup] no SDN setup required")
return false, nil
}
glog.V(5).Infof("[SDN setup] full SDN setup required")
mtuStr := fmt.Sprint(mtu)
itx := ipcmd.NewTransaction(LBR)
itx.SetLink("down")
itx.IgnoreError()
itx.DeleteLink()
itx.IgnoreError()
itx.AddLink("type", "bridge")
itx.AddAddress(gwCIDR)
itx.SetLink("up")
err = itx.EndTransaction()
if err != nil {
glog.Errorf("Failed to configure docker bridge: %v", err)
return false, err
}
defer deleteLocalSubnetRoute(LBR, localSubnetCIDR)
glog.V(5).Infof("[SDN setup] docker setup %s mtu %s", LBR, mtuStr)
out, err := exec.Command("openshift-sdn-docker-setup.sh", LBR, mtuStr).CombinedOutput()
if err != nil {
glog.Errorf("Failed to configure docker networking: %v\n%s", err, out)
return false, err
} else {
glog.V(5).Infof("[SDN setup] docker setup success:\n%s", out)
}
config := fmt.Sprintf("export OPENSHIFT_CLUSTER_SUBNET=%s", clusterNetworkCIDR)
err = ioutil.WriteFile("/run/openshift-sdn/config.env", []byte(config), 0644)
if err != nil {
return false, err
}
itx = ipcmd.NewTransaction(VLINUXBR)
itx.DeleteLink()
itx.IgnoreError()
itx.AddLink("mtu", mtuStr, "type", "veth", "peer", "name", VOVSBR, "mtu", mtuStr)
itx.SetLink("up")
itx.SetLink("txqueuelen", "0")
err = itx.EndTransaction()
if err != nil {
return false, err
}
itx = ipcmd.NewTransaction(VOVSBR)
itx.SetLink("up")
itx.SetLink("txqueuelen", "0")
err = itx.EndTransaction()
if err != nil {
return false, err
}
itx = ipcmd.NewTransaction(LBR)
itx.AddSlave(VLINUXBR)
err = itx.EndTransaction()
if err != nil {
return false, err
}
otx := ovs.NewTransaction(BR)
otx.AddBridge("fail-mode=secure", "protocols=OpenFlow13")
otx.AddPort(VXLAN, 1, "type=vxlan", `options:remote_ip="flow"`, `options:key="flow"`)
otx.AddPort(TUN, 2, "type=internal")
otx.AddPort(VOVSBR, 3)
// Table 0; VXLAN filtering; the first rule sends un-tunnelled packets
// to table 1. Additional per-node rules are filled in by controller.go
otx.AddFlow("table=0, tun_src=0.0.0.0, actions=goto_table:1")
// eg, "table=0, tun_src=${remote_node}, actions=goto_table:1"
// Table 1; learn MAC addresses and continue with table 2
otx.AddFlow("table=1, actions=learn(table=9, priority=200, hard_timeout=900, NXM_OF_ETH_DST[]=NXM_OF_ETH_SRC[], load:NXM_NX_TUN_IPV4_SRC[]->NXM_NX_TUN_IPV4_DST[], output:NXM_OF_IN_PORT[]), goto_table:2")
// Table 2; initial dispatch
otx.AddFlow("table=2, priority=200, arp, actions=goto_table:9")
otx.AddFlow("table=2, priority=100, in_port=1, actions=goto_table:3") // vxlan0
otx.AddFlow("table=2, priority=100, in_port=2, actions=goto_table:6") // tun0
otx.AddFlow("table=2, priority=100, in_port=3, actions=goto_table:6") // vovsbr
otx.AddFlow("table=2, priority=0, actions=goto_table:4") // container
// Table 3; incoming from vxlan
otx.AddFlow("table=3, priority=200, ip, nw_dst=%s, actions=output:2", localSubnetGateway)
if c.multitenant {
otx.AddFlow("table=3, priority=100, ip, nw_dst=%s, actions=move:NXM_NX_TUN_ID[0..31]->NXM_NX_REG0[], goto_table:6", localSubnetCIDR)
} else {
otx.AddFlow("table=3, priority=100, ip, nw_dst=%s, actions=goto_table:9", localSubnetCIDR)
}
// Table 4; incoming from container; filled in by openshift-sdn-ovs
// eg, single-tenant: "table=4, priority=100, in_port=${ovs_port}, ip, nw_src=${ipaddr}, goto_table:6"
// multitenant: "table=4, priority=100, in_port=${ovs_port}, ip, nw_src=${ipaddr}, actions=load:${tenant_id}->NXM_NX_REG0[], goto_table:5"
// Table 5; service isolation; mostly filled in by AddServiceOFRules()
if c.multitenant {
otx.AddFlow("table=5, priority=200, reg0=0, ip, nw_dst=%s, actions=output:2", servicesNetworkCIDR)
// eg, "table=5, priority=200, ${service_proto}, nw_dst=${service_ip}, tp_dst=${service_port}, actions=output:2"
otx.AddFlow("table=5, priority=100, ip, nw_dst=%s, actions=drop", servicesNetworkCIDR)
otx.AddFlow("table=5, priority=0, actions=goto_table:6")
}
// Table 6; general routing
otx.AddFlow("table=6, priority=200, ip, nw_dst=%s, actions=output:2", localSubnetGateway)
if c.multitenant {
otx.AddFlow("table=6, priority=175, ip, reg0=0, nw_dst=%s, actions=goto_table:9", localSubnetCIDR)
otx.AddFlow("table=6, priority=150, ip, nw_dst=%s, actions=goto_table:7", localSubnetCIDR)
} else {
otx.AddFlow("table=6, priority=150, ip, nw_dst=%s, actions=goto_table:9", localSubnetCIDR)
}
otx.AddFlow("table=6, priority=100, ip, nw_dst=%s, actions=goto_table:8", clusterNetworkCIDR)
otx.AddFlow("table=6, priority=0, ip, actions=output:2")
// Table 7; to local container with isolation; filled in by openshift-sdn-ovs
// eg, "table=7, priority=100, ip, nw_dst=${ipaddr}, reg0=${tenant_id}, actions=output:${ovs_port}"
// Table 8; to remote container; filled in by AddOFRules()
// eg, "table=8, priority=100, ip, nw_dst=${remote_subnet_cidr}, actions=move:NXM_NX_REG0[]->NXM_NX_TUN_ID[0..31], set_field:${remote_node_ip}->tun_dst,output:1"
// Table 9; MAC dispatch / ARP, filled in by Table 1's learn() rule
// and with per-node vxlan ARP rules by AddOFRules()
otx.AddFlow("table=9, priority=0, arp, actions=flood")
// eg, "table=9, priority=100, arp, nw_dst=${remote_subnet_cidr}, actions=move:NXM_NX_REG0[]->NXM_NX_TUN_ID[0..31], set_field:${remote_node_ip}->tun_dst,output:1"
err = otx.EndTransaction()
if err != nil {
return false, err
}
itx = ipcmd.NewTransaction(TUN)
itx.AddAddress(gwCIDR)
defer deleteLocalSubnetRoute(TUN, localSubnetCIDR)
itx.SetLink("mtu", mtuStr)
itx.SetLink("up")
itx.AddRoute(clusterNetworkCIDR, "proto", "kernel", "scope", "link")
itx.AddRoute(servicesNetworkCIDR)
err = itx.EndTransaction()
if err != nil {
return false, err
}
// Clean up docker0 since docker won't
itx = ipcmd.NewTransaction("docker0")
itx.SetLink("down")
itx.IgnoreError()
itx.DeleteLink()
itx.IgnoreError()
_ = itx.EndTransaction()
// Disable iptables for linux bridges (and in particular lbr0), ignoring errors.
// (This has to have been performed in advance for docker-in-docker deployments,
// since this will fail there).
_, _ = exec.Command("modprobe", "br_netfilter").CombinedOutput()
err = sysctl.SetSysctl("net/bridge/bridge-nf-call-iptables", 0)
if err != nil {
glog.Warningf("Could not set net.bridge.bridge-nf-call-iptables sysctl: %s", err)
} else {
glog.V(5).Infof("[SDN setup] set net.bridge.bridge-nf-call-iptables to 0")
}
// Enable IP forwarding for ipv4 packets
err = sysctl.SetSysctl("net/ipv4/ip_forward", 1)
if err != nil {
return false, fmt.Errorf("Could not enable IPv4 forwarding: %s", err)
}
err = sysctl.SetSysctl(fmt.Sprintf("net/ipv4/conf/%s/forwarding", TUN), 1)
if err != nil {
return false, fmt.Errorf("Could not enable IPv4 forwarding on %s: %s", TUN, err)
}
return true, nil
}
func (realConntracker) SetTCPEstablishedTimeout(seconds int) error {
glog.Infof("Setting nf_conntrack_tcp_timeout_established to %d", seconds)
return sysctl.SetSysctl("net/netfilter/nf_conntrack_tcp_timeout_established", seconds)
}
func (plugin *OsdnNode) SetupSDN() (bool, error) {
localSubnetCIDR, err := plugin.getLocalSubnet()
if err != nil {
return false, err
}
clusterNetworkCIDR := plugin.networkInfo.ClusterNetwork.String()
serviceNetworkCIDR := plugin.networkInfo.ServiceNetwork.String()
_, ipnet, err := net.ParseCIDR(localSubnetCIDR)
localSubnetMaskLength, _ := ipnet.Mask.Size()
localSubnetGateway := netutils.GenerateDefaultGateway(ipnet).String()
glog.V(5).Infof("[SDN setup] node pod subnet %s gateway %s", ipnet.String(), localSubnetGateway)
gwCIDR := fmt.Sprintf("%s/%d", localSubnetGateway, localSubnetMaskLength)
if plugin.alreadySetUp(gwCIDR, clusterNetworkCIDR) {
glog.V(5).Infof("[SDN setup] no SDN setup required")
return false, nil
}
glog.V(5).Infof("[SDN setup] full SDN setup required")
mtuStr := fmt.Sprint(plugin.mtu)
exec := kexec.New()
itx := ipcmd.NewTransaction(exec, LBR)
itx.SetLink("down")
itx.IgnoreError()
itx.DeleteLink()
itx.IgnoreError()
itx.AddLink("type", "bridge")
itx.AddAddress(gwCIDR)
itx.SetLink("up")
err = itx.EndTransaction()
if err != nil {
glog.Errorf("Failed to configure docker bridge: %v", err)
return false, err
}
defer deleteLocalSubnetRoute(LBR, localSubnetCIDR)
glog.V(5).Infof("[SDN setup] docker setup %s mtu %s", LBR, mtuStr)
out, err := exec.Command("openshift-sdn-docker-setup.sh", LBR, mtuStr).CombinedOutput()
if err != nil {
glog.Errorf("Failed to configure docker networking: %v\n%s", err, out)
return false, err
} else {
glog.V(5).Infof("[SDN setup] docker setup success:\n%s", out)
}
config := fmt.Sprintf("export OPENSHIFT_CLUSTER_SUBNET=%s", clusterNetworkCIDR)
err = ioutil.WriteFile("/run/openshift-sdn/config.env", []byte(config), 0644)
if err != nil {
return false, err
}
itx = ipcmd.NewTransaction(exec, VLINUXBR)
itx.DeleteLink()
itx.IgnoreError()
itx.AddLink("mtu", mtuStr, "type", "veth", "peer", "name", VOVSBR, "mtu", mtuStr)
itx.SetLink("up")
itx.SetLink("txqueuelen", "0")
err = itx.EndTransaction()
if err != nil {
return false, err
}
itx = ipcmd.NewTransaction(exec, VOVSBR)
itx.SetLink("up")
itx.SetLink("txqueuelen", "0")
err = itx.EndTransaction()
if err != nil {
return false, err
}
itx = ipcmd.NewTransaction(exec, LBR)
itx.AddSlave(VLINUXBR)
err = itx.EndTransaction()
if err != nil {
return false, err
}
err = plugin.ovs.AddBridge("fail-mode=secure", "protocols=OpenFlow13")
if err != nil {
return false, err
}
_ = plugin.ovs.DeletePort(VXLAN)
_, err = plugin.ovs.AddPort(VXLAN, 1, "type=vxlan", `options:remote_ip="flow"`, `options:key="flow"`)
if err != nil {
return false, err
}
_ = plugin.ovs.DeletePort(TUN)
_, err = plugin.ovs.AddPort(TUN, 2, "type=internal")
if err != nil {
return false, err
}
_, err = plugin.ovs.AddPort(VOVSBR, 3)
if err != nil {
return false, err
}
otx := plugin.ovs.NewTransaction()
// Table 0: initial dispatch based on in_port
// vxlan0
otx.AddFlow("table=0, priority=200, in_port=1, arp, nw_src=%s, nw_dst=%s, actions=move:NXM_NX_TUN_ID[0..31]->NXM_NX_REG0[],goto_table:1", clusterNetworkCIDR, localSubnetCIDR)
otx.AddFlow("table=0, priority=200, in_port=1, ip, nw_src=%s, nw_dst=%s, actions=move:NXM_NX_TUN_ID[0..31]->NXM_NX_REG0[],goto_table:1", clusterNetworkCIDR, localSubnetCIDR)
otx.AddFlow("table=0, priority=150, in_port=1, actions=drop")
// tun0
otx.AddFlow("table=0, priority=200, in_port=2, arp, nw_src=%s, nw_dst=%s, actions=goto_table:5", localSubnetGateway, clusterNetworkCIDR)
otx.AddFlow("table=0, priority=200, in_port=2, ip, actions=goto_table:5")
otx.AddFlow("table=0, priority=150, in_port=2, actions=drop")
// vovsbr
otx.AddFlow("table=0, priority=200, in_port=3, arp, nw_src=%s, actions=goto_table:5", localSubnetCIDR)
otx.AddFlow("table=0, priority=200, in_port=3, ip, nw_src=%s, actions=goto_table:5", localSubnetCIDR)
otx.AddFlow("table=0, priority=150, in_port=3, actions=drop")
// else, from a container
otx.AddFlow("table=0, priority=100, arp, actions=goto_table:2")
otx.AddFlow("table=0, priority=100, ip, actions=goto_table:2")
otx.AddFlow("table=0, priority=0, actions=drop")
// Table 1: VXLAN ingress filtering; filled in by AddHostSubnetRules()
// eg, "table=1, priority=100, tun_src=${remote_node_ip}, actions=goto_table:5"
otx.AddFlow("table=1, priority=0, actions=drop")
// Table 2: from OpenShift container; validate IP/MAC, assign tenant-id; filled in by openshift-sdn-ovs
// eg, "table=2, priority=100, in_port=${ovs_port}, arp, nw_src=${ipaddr}, arp_sha=${macaddr}, actions=load:${tenant_id}->NXM_NX_REG0[], goto_table:5"
// "table=2, priority=100, in_port=${ovs_port}, ip, nw_src=${ipaddr}, actions=load:${tenant_id}->NXM_NX_REG0[], goto_table:3"
// (${tenant_id} is always 0 for single-tenant)
otx.AddFlow("table=2, priority=0, actions=drop")
// Table 3: from OpenShift container; service vs non-service
otx.AddFlow("table=3, priority=100, ip, nw_dst=%s, actions=goto_table:4", serviceNetworkCIDR)
otx.AddFlow("table=3, priority=0, actions=goto_table:5")
// Table 4: from OpenShift container; service dispatch; filled in by AddServiceRules()
otx.AddFlow("table=4, priority=200, reg0=0, actions=output:2")
// eg, "table=4, priority=100, reg0=${tenant_id}, ${service_proto}, nw_dst=${service_ip}, tp_dst=${service_port}, actions=output:2"
otx.AddFlow("table=4, priority=0, actions=drop")
// Table 5: general routing
otx.AddFlow("table=5, priority=300, arp, nw_dst=%s, actions=output:2", localSubnetGateway)
otx.AddFlow("table=5, priority=300, ip, nw_dst=%s, actions=output:2", localSubnetGateway)
otx.AddFlow("table=5, priority=200, arp, nw_dst=%s, actions=goto_table:6", localSubnetCIDR)
otx.AddFlow("table=5, priority=200, ip, nw_dst=%s, actions=goto_table:7", localSubnetCIDR)
otx.AddFlow("table=5, priority=100, arp, nw_dst=%s, actions=goto_table:8", clusterNetworkCIDR)
otx.AddFlow("table=5, priority=100, ip, nw_dst=%s, actions=goto_table:8", clusterNetworkCIDR)
otx.AddFlow("table=5, priority=0, ip, actions=goto_table:9")
otx.AddFlow("table=5, priority=0, arp, actions=drop")
// Table 6: ARP to container, filled in by openshift-sdn-ovs
// eg, "table=6, priority=100, arp, nw_dst=${container_ip}, actions=output:${ovs_port}"
otx.AddFlow("table=6, priority=0, actions=output:3")
// Table 7: IP to container; filled in by openshift-sdn-ovs
// eg, "table=7, priority=100, reg0=0, ip, nw_dst=${ipaddr}, actions=output:${ovs_port}"
// eg, "table=7, priority=100, reg0=${tenant_id}, ip, nw_dst=${ipaddr}, actions=output:${ovs_port}"
otx.AddFlow("table=7, priority=0, actions=output:3")
// Table 8: to remote container; filled in by AddHostSubnetRules()
// eg, "table=8, priority=100, arp, nw_dst=${remote_subnet_cidr}, actions=move:NXM_NX_REG0[]->NXM_NX_TUN_ID[0..31], set_field:${remote_node_ip}->tun_dst,output:1"
// eg, "table=8, priority=100, ip, nw_dst=${remote_subnet_cidr}, actions=move:NXM_NX_REG0[]->NXM_NX_TUN_ID[0..31], set_field:${remote_node_ip}->tun_dst,output:1"
otx.AddFlow("table=8, priority=0, actions=drop")
// Table 9: egress network policy dispatch; edited by updateEgressNetworkPolicyRules()
// eg, "table=9, reg0=${tenant_id}, priority=2, ip, nw_dst=${external_cidr}, actions=drop
otx.AddFlow("table=9, priority=0, actions=output:2")
err = otx.EndTransaction()
if err != nil {
return false, err
}
itx = ipcmd.NewTransaction(exec, TUN)
itx.AddAddress(gwCIDR)
defer deleteLocalSubnetRoute(TUN, localSubnetCIDR)
itx.SetLink("mtu", mtuStr)
itx.SetLink("up")
itx.AddRoute(clusterNetworkCIDR, "proto", "kernel", "scope", "link")
itx.AddRoute(serviceNetworkCIDR)
err = itx.EndTransaction()
if err != nil {
return false, err
}
// Clean up docker0 since docker won't
itx = ipcmd.NewTransaction(exec, "docker0")
itx.SetLink("down")
itx.IgnoreError()
itx.DeleteLink()
itx.IgnoreError()
_ = itx.EndTransaction()
sysctl := sysctl.New()
// Disable iptables for linux bridges (and in particular lbr0), ignoring errors.
// (This has to have been performed in advance for docker-in-docker deployments,
// since this will fail there).
_, _ = exec.Command("modprobe", "br_netfilter").CombinedOutput()
err = sysctl.SetSysctl("net/bridge/bridge-nf-call-iptables", 0)
if err != nil {
glog.Warningf("Could not set net.bridge.bridge-nf-call-iptables sysctl: %s", err)
} else {
glog.V(5).Infof("[SDN setup] set net.bridge.bridge-nf-call-iptables to 0")
}
// Enable IP forwarding for ipv4 packets
err = sysctl.SetSysctl("net/ipv4/ip_forward", 1)
if err != nil {
return false, fmt.Errorf("Could not enable IPv4 forwarding: %s", err)
}
err = sysctl.SetSysctl(fmt.Sprintf("net/ipv4/conf/%s/forwarding", TUN), 1)
if err != nil {
return false, fmt.Errorf("Could not enable IPv4 forwarding on %s: %s", TUN, err)
}
// Table 253: rule version; note action is hex bytes separated by '.'
otx = plugin.ovs.NewTransaction()
pluginVersion := getPluginVersion(plugin.multitenant)
otx.AddFlow("%s, %s%s.%s", VERSION_TABLE, VERSION_ACTION, pluginVersion[0], pluginVersion[1])
err = otx.EndTransaction()
if err != nil {
return false, err
}
return true, nil
}