[IPv6] adjust MTU for IPv6 overhead #1305

lzhecheng · 2020-09-25T08:57:03Z

If MTU is too large for IPv6 environment, large packet with overhead
IPv6ExtraOverhead, 20 is from observation of IPv4 and IPv6 packets under same situation.

antrea-bot · 2020-09-25T08:57:16Z

Thanks for your PR.
Unit tests and code linters are run automatically every time the PR is updated.
E2e, conformance and network policy tests can only be triggered by a member of the vmware-tanzu organization. Regular contributors to the project should join the org.

The following commands are available:

/test-e2e: to trigger e2e tests.
/skip-e2e: to skip e2e tests.
/test-conformance: to trigger conformance tests.
/skip-conformance: to skip conformance tests.
/test-whole-conformance: to trigger all conformance tests on linux.
/skip-whole-conformance: to skip all conformance tests on linux.
/test-networkpolicy: to trigger networkpolicy tests.
/skip-networkpolicy: to skip networkpolicy tests.
/test-windows-conformance: to trigger windows conformance tests.
/skip-windows-conformance: to skip windows conformance tests.
/test-windows-networkpolicy: to trigger windows networkpolicy tests.
/skip-windows-networkpolicy: to skip windows networkpolicy tests.
/test-hw-offload: to trigger ovs hardware offload test.
/skip-hw-offload: to skip ovs hardware offload test.
/test-all: to trigger all tests (except whole conformance).
/skip-all: to skip all tests (except whole conformance).

codecov-commenter · 2020-09-25T09:01:12Z

Codecov Report

Merging #1305 into ipv6 will decrease coverage by 0.03%.
The diff coverage is 16.66%.

@@            Coverage Diff             @@
##             ipv6    #1305      +/-   ##
==========================================
- Coverage   57.91%   57.88%   -0.04%     
==========================================
  Files         155      155              
  Lines       13032    13034       +2     
==========================================
- Hits         7548     7545       -3     
- Misses       4681     4684       +3     
- Partials      803      805       +2

Flag	Coverage Δ
#kind-e2e-tests	`46.23% <16.66%> (+0.04%)`	⬆️
#unit-tests	`41.53% <0.00%> (-0.08%)`	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
pkg/agent/config/node_config.go	`100.00% <ø> (ø)`
pkg/agent/openflow/client.go	`41.86% <ø> (ø)`
pkg/agent/agent.go	`49.32% <16.66%> (-0.27%)`	⬇️
pkg/apiserver/certificate/certificate.go	`71.05% <0.00%> (-6.58%)`	⬇️
pkg/apiserver/storage/ram/watch.go	`90.38% <0.00%> (-3.85%)`	⬇️
pkg/monitor/controller.go	`57.35% <0.00%> (+5.88%)`	⬆️

lzhecheng · 2020-09-25T09:08:57Z

/test-ipv6-only-conformance
/test-ipv6-only-e2e

lzhecheng · 2020-09-25T09:26:40Z

/test-windows-conformance

lzhecheng · 2020-09-25T09:29:25Z

/test-ipv6-only-conformance
/test-ipv6-only-e2e

build/yamls/antrea-aks.yml

lzhecheng · 2020-09-28T02:22:46Z

/test-ipv6-only-e2e

jianjuns

Question - is it possible we use IPv4 tunnel when Pod CIDR is IPv6? @wenyingd

build/yamls/antrea-aks.yml

wenyingd · 2020-09-29T01:35:03Z

Question - is it possible we use IPv4 tunnel when Pod CIDR is IPv6? @wenyingd

In theory, it is possible. We use Node's Internal Address to set tunnel header. If there is a Dual-Stack cluster, and the Node reports its address using an IPv6, then for the Pod's traffic in IPv4 should be transmitted with IPv6 tunnel.

jianjuns · 2020-09-29T04:57:21Z

Question - is it possible we use IPv4 tunnel when Pod CIDR is IPv6? @wenyingd

In theory, it is possible. We use Node's Internal Address to set tunnel header. If there is a Dual-Stack cluster, and the Node reports its address using an IPv6, then for the Pod's traffic in IPv4 should be transmitted with IPv6 tunnel.

Got it. Maybe good to add comments somewhere about this behavior (why IPv6 or v4 tunnel is decided by Pod CDIR is v4 or v6).
@lzhecheng

If Antrea MTU is too large in IPv6 environment, large packet with overhead exceeds node MTU cannot be transmitted successfully across nodes. IPv6ExtraOverhead, 20 is from observation of IPv4 and IPv6 packets under same situation.

lzhecheng · 2020-09-29T08:25:23Z

/test-all
/test-ipv6-e2e
/test-ipv6-only-e2e

lzhecheng · 2020-09-29T08:28:35Z

Question - is it possible we use IPv4 tunnel when Pod CIDR is IPv6? @wenyingd

In theory, it is possible. We use Node's Internal Address to set tunnel header. If there is a Dual-Stack cluster, and the Node reports its address using an IPv6, then for the Pod's traffic in IPv4 should be transmitted with IPv6 tunnel.

Got it. Maybe good to add comments somewhere about this behavior (why IPv6 or v4 tunnel is decided by Pod CDIR is v4 or v6).
@lzhecheng

Got it. Updated.

lzhecheng · 2020-09-29T08:37:09Z

/test-ipv6-e2e
/test-ipv6-only-e2e

lzhecheng · 2020-09-29T09:27:08Z

/test-windows-networkpolicy

lzhecheng · 2020-09-29T09:51:05Z

/test-ipv6-conformance
/test-ipv6-only-conformance

lzhecheng · 2020-09-29T10:44:06Z

/test-ipv6-conformance
/test-ipv6-only-conformance

wenyingd · 2020-09-29T12:03:38Z

pkg/agent/agent.go

@@ -730,6 +729,9 @@ func (i *Initializer) getNodeMTU(localIntf *net.Interface) (int, error) {
 	if i.networkConfig.EnableIPSecTunnel {
 		mtu -= config.IpsecESPOverhead
 	}
+	if i.nodeConfig.PodIPv6CIDR != nil {


I am thinking if IPv6 extra cost is in the tunnel header, shall we check the address family of Node's Internal Address instead of the PodCIDR's?

Make sense. Let's created a new PR to fix this.

This reverts commit 044c476.

If Antrea MTU is too large in IPv6 environment, large packet with overhead exceeds node MTU cannot be transmitted successfully across nodes. IPv6ExtraOverhead, 20 is from observation of IPv4 and IPv6 packets under same situation.

* Enable verbose logging through generate-manifest.sh (#1142) This patch enables verbose logging for antrea-agent and antrea-controller when generate manifest. This will help in troubleshooting with increased log level. * Bug in PR#1142 (#1248) manifest generation failing in release mode. Fixed the bug. * [IPv6] Consume Node.Spec.CIDRs to support dual-stack configuration (#971) 1. Consume Node.Spec.CIDRs to support IPv4/IPv6 dual-stack Pod Subnets 2. Change NodeConfig.PodCIDR as a slice 3. Change GatewayConfig.IP as a slice to support multiple addresses for antrea-gw0 4. Change InterfaceConfig.IP as a slice to support multiple address for a Pod * [IPv6] Change openflow pipeline for L2 Pod networking (#1040) 1. Add a new table named IPv6 to handle IPv6 ND Solicitation, ND advertisement and IPv6 Multicast traffic. 2. Add flows in openflow tables (spoofGuardTable, IPv6, conntrackTable, conntrackStateTable, conntrackCommitTable, L2ForwardingOutTable) for handling IPv6 L2 Pod networking. * [IPv6] Change host-local IPAM configuration for IPv6 (#1039) 1. Add new field Ranges in IPAMConfig for allocating both IPv4 and IPv6 addresses. 2. Populate subnet and gateway for both IPv4 range and IPv6 range. * [IPv6] Use separate fields for IPv4 and IPv6 in GatewayConfig (#1111) Replace IP slice in GatewayConfig with separate IPv4 and IPv6 fields. * [IPv6] Implement L3 connectivity for IPv6 traffic (#1011) 1. Use IPv6 in iptables and ipset configuration. 2. Identifiy IPv6 address and configure in OpenFlow. 3. Use Node Internal address for tunnel. * [IPv6] Handle Spec.PodCIDR with IPv6 CIDR (#1151) For IPv6 single stack case, node.Spec.PodCIDR is configured with IPv6 CIDR. This patch handles the case and sets nodeConfig.PodIPv6CIDR with parsed CIDR. * [IPv6] Add support for IPv6 address in antctl and agent's apiserver (#1118) * [IPv6] Add support for IPv6 address in antctl and agent's apiserver 1. Support using IPv6 address in OVS tracing. 2. Support displaying Node's and Pod's IPv6 address in agent apiserver. Co-authored-by: Zhecheng Li <zhechel1@uci.edu> * [IPv6] Support IPv6 in e2e (#1129) * [IPv6] Display dual stack NodeSubnet in antrea-octant-plugin (#1156) NodeSubnet can have two values in dual stack case and this patch enhances octant-plugin to show both subnets. * [IPv6] Handle dual stack NodeSubnet for monitoring CRD (#1182) 1. Rename NodeSubnet to NodeSubnets for AntreaAgentInfo. 2. Make a new string slice for dual stack node subnet instead of appending agentInfo.NodeSubnets directly to avoid duplicate CIDRs. * [IPv6][e2e] Fix testDeletePod (#1193) On a dual-stack cluster, podInterfaces[0].IP returns "[ipv4-address], [ipv6-address]". Current implementation doesn't distingush two. * [IPv6] Collect service CIDR in e2e * [IPv6] Add support for dual-stack when using kube-proxy for Service (#1200) 1. Add a config item for IPv6 Serivce CIDR if using kube-proxy to provide Service functions. 2. Output IPv6 traffic from host gateway if its destination is a Service address. 3. Use ct_mark to identify Service traffic and output the reply packet to the host gateway to ensure the DNAT processing in iptables. * [IPv6] Extend e2e tests for dual-stack (#1192) 1. Extend generic functions "podWaitForIP" to return all assigned IPs of a given Pod. 2. Validate each IP address in the cluster's network CIDR 3. Use each valid IP to check connectivity. 4. Use each valid IP to execute tests. * [IPv6] E2e bug fixes (#1311) 1. No -6 option in busybox nc So, no need to distinguish if it is an IPv6 environment for runNetcatCommandFromTestPod() nc BusyBox v1.31.1 (2019-10-28 18:40:01 UTC) multi-call binary. Usage: nc [OPTIONS] HOST PORT - connect nc [OPTIONS] -l -p PORT [HOST] [PORT] - listen -e PROG Run PROG after connect (must be last) -l Listen mode, for inbound connects -lk With -e, provides persistent server -p PORT Local port -s ADDR Local address -w SEC Timeout for connects and final net reads -i SEC Delay interval for lines sent -n Don't do DNS resolution -u UDP mode -v Verbose -o FILE Hex dump traffic -z Zero-I/O mode (scanning) 2. testCert * IPv6 address should be in "[]" * [IPv6] Fix TestReconcileGatewayRoutesOnStartup failure (#1313) Use "ip -6 route" for IPv6 network. * [IPv6] adjust MTU for IPv6 overhead (#1305) If Antrea MTU is too large in IPv6 environment, large packet with overhead exceeds node MTU cannot be transmitted successfully across nodes. IPv6ExtraOverhead, 20 is from observation of IPv4 and IPv6 packets under same situation. * [IPv6] Fix MTU config (#1317) Use Node's internal address to decide if extra IPv6 overhead is needed. * [IPv6] Skip IPsec e2e test (#1373) * With OVS v2.14.0, IPsec in IPv6 envinronment is not supported. * More user-friendly output for PodIPs. From: Retrieved all Pod IPs: map[test-pod-0-upgp1ung:0xc000708960 test-pod-1-pbva9007:0xc0006ec8a0] To: Retrieved all Pod IPs: map[test-pod-0-mudzj847:IPv6: fd74:ca9b:172:16::4, IP strings: fd74:ca9b:172:16::4 test-pod-1-apcmyd30:IPv6: fd74:ca9b:172:16:1::3c, IP strings: fd74:ca9b:172:16:1::3c] * [IPv6] Add 2 Network Policy tests (#1399) 2 upstream Network Policy tests didn't consider netmask for IPv6, this patch is to add correct tests. When bug is fixed in latest release, these 2 tests can be deleted. Kubernetes PR: kubernetes/kubernetes#93583 2 testcases: https://github.com/kubernetes/kubernetes/blob/v1.20.0-alpha.0/test/e2e/network/network_policy.go#L1365 https://github.com/kubernetes/kubernetes/blob/v1.20.0-alpha.0/test/e2e/network/network_policy.go#L1444 * Skip 2 Network Policy testcases before Network Policy IPv6 is supported (#1460) * [IPv6] Fix after rebasing * format code * fix TestPodTrafficShaping * fix TestIPv6RoutesAndNeighbors * [IPv6] Fix issues (#1496) * unit test * manifest * [IPv6] Skip TestAntctlProxy for IPv6 (#1498) * [IPv6] Add IPv6 support for NetworkPolicy 1. Add enhancement in Antrea Controller and Agent to support NetworkPolicy in IPv6. 2. Optimize test cases to support IPv6 3. Use regex in CRD to validate IPv4 or IPv6 string 4. Add TestEgressToServerInCIDRBlock and TestEgressToServerInCIDRBlockWithException 5. networkpolicy_controller.go: PodIPs includes PodIP * [IPv6] Fix issues * remove Github Actions integration test, Jenkins: jenkins-integration -> Integration tests * go fmt * add FlowProtocl() to interface Flow * remove extra lines when rebasing for an octant commit * TestIPv6RoutesAndNeighbors: routeClient.Initialize Co-authored-by: srikartati <stati@vmware.com> Co-authored-by: Wenying Dong <wenyingd@vmware.com> Co-authored-by: Mengdie Song <songm@vmware.com> Co-authored-by: Zhecheng Li <zhechel1@uci.edu>

lzhecheng requested review from wenyingd and jianjuns September 25, 2020 08:57

vmwclabot added the cla-not-required label Sep 25, 2020

wenyingd force-pushed the ipv6 branch from 15eea1d to cf688b3 Compare September 25, 2020 12:26

jianjuns reviewed Sep 25, 2020

View reviewed changes

build/yamls/antrea-aks.yml Outdated Show resolved Hide resolved

lzhecheng force-pushed the ipv6-mtu branch 2 times, most recently from 29a760c to 98febd3 Compare September 28, 2020 02:19

lzhecheng requested a review from jianjuns September 28, 2020 02:40

jianjuns reviewed Sep 28, 2020

View reviewed changes

build/yamls/antrea-aks.yml Outdated Show resolved Hide resolved

lzhecheng force-pushed the ipv6-mtu branch from 98febd3 to d552981 Compare September 29, 2020 01:12

lzhecheng requested a review from jianjuns September 29, 2020 01:14

jianjuns approved these changes Sep 29, 2020

View reviewed changes

wenyingd force-pushed the ipv6 branch from c5577c2 to ccad6ea Compare September 29, 2020 07:06

[IPv6] adjust MTU for IPv6 overhead

61fd805

If Antrea MTU is too large in IPv6 environment, large packet with overhead exceeds node MTU cannot be transmitted successfully across nodes. IPv6ExtraOverhead, 20 is from observation of IPv4 and IPv6 packets under same situation.

lzhecheng force-pushed the ipv6-mtu branch from d552981 to 61fd805 Compare September 29, 2020 08:24

lzhecheng merged commit 044c476 into antrea-io:ipv6 Sep 29, 2020

lzhecheng deleted the ipv6-mtu branch September 29, 2020 12:02

wenyingd reviewed Sep 29, 2020

View reviewed changes

lzhecheng restored the ipv6-mtu branch September 29, 2020 12:12

lzhecheng pushed a commit that referenced this pull request Sep 29, 2020

Revert "[IPv6] adjust MTU for IPv6 overhead (#1305)"

3c0fab2

This reverts commit 044c476.

lzhecheng deleted the ipv6-mtu branch September 29, 2020 12:16

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[IPv6] adjust MTU for IPv6 overhead #1305

[IPv6] adjust MTU for IPv6 overhead #1305

lzhecheng commented Sep 25, 2020 •

edited

Loading

antrea-bot commented Sep 25, 2020

codecov-commenter commented Sep 25, 2020 •

edited

Loading

lzhecheng commented Sep 25, 2020

lzhecheng commented Sep 25, 2020

lzhecheng commented Sep 25, 2020

lzhecheng commented Sep 28, 2020

jianjuns left a comment

wenyingd commented Sep 29, 2020

jianjuns commented Sep 29, 2020

lzhecheng commented Sep 29, 2020

lzhecheng commented Sep 29, 2020

lzhecheng commented Sep 29, 2020

lzhecheng commented Sep 29, 2020

lzhecheng commented Sep 29, 2020

lzhecheng commented Sep 29, 2020

wenyingd Sep 29, 2020

lzhecheng Sep 29, 2020

[IPv6] adjust MTU for IPv6 overhead #1305

[IPv6] adjust MTU for IPv6 overhead #1305

Conversation

lzhecheng commented Sep 25, 2020 • edited Loading

antrea-bot commented Sep 25, 2020

codecov-commenter commented Sep 25, 2020 • edited Loading

Codecov Report

lzhecheng commented Sep 25, 2020

lzhecheng commented Sep 25, 2020

lzhecheng commented Sep 25, 2020

lzhecheng commented Sep 28, 2020

jianjuns left a comment

Choose a reason for hiding this comment

wenyingd commented Sep 29, 2020

jianjuns commented Sep 29, 2020

lzhecheng commented Sep 29, 2020

lzhecheng commented Sep 29, 2020

lzhecheng commented Sep 29, 2020

lzhecheng commented Sep 29, 2020

lzhecheng commented Sep 29, 2020

lzhecheng commented Sep 29, 2020

wenyingd Sep 29, 2020

Choose a reason for hiding this comment

lzhecheng Sep 29, 2020

Choose a reason for hiding this comment

lzhecheng commented Sep 25, 2020 •

edited

Loading

codecov-commenter commented Sep 25, 2020 •

edited

Loading