Problem:
After multi clusters registered to a unfied control panel, the pod IP reachability between pods and services across clusters are required for HA and distributed application scheduling case, the submariner is the best choice against cilium mash and istio multi cluster mesh, since it has little dependence and better align with registered cluster deploy topology
Scope:
- (p0) The operator is able to install
broker and agent components by CR
- (p0) The operatoer should conver globalnet case when cluster cidr overlap
- (p1)The operator shoud leverage [cloud-prepare](https://github.com/submariner-io/cloud-prepare to prepare cluster env according to to different hosted k8s vonder such ACK, IKS, TKE
- p0 is mandatory
- p1 is nice to have
Case result:
| broker k8s |
mannaged k8s |
pod/pod connectivity |
service discovery |
CNI version |
deployment mode |
kube-proxy mode |
note |
| 1.15.7 |
1.15.7 |
ok (pod cidr: xxxxxxxxxxxxxxxxxxxxxxxxxxx) |
ok (service cidr: xxxxxxxxxxxxxxxxxxxxxxxxxxx) |
flannel v0.14.0 |
On-Premise |
iptables |
subctl v0.6.0-dev |
| 1.16.7 |
1.16.7 |
ok (pod cidr: 10.44.0.0/16; 10.144.0.0/16) |
ok (service cidr: 10.45.0.0/16; 10.145.0.0/16) |
calicoctl v3.18.4 |
On-Premise |
iptables |
subctl v0.9.0-dev |
| 1.19.1 |
1.19.1 |
ok (pod cidr: 10.44.0.0/16; 10.144.0.0/16) |
ok (service cidr: 10.45.0.0/16; 10.145.0.0/16) |
calicoctl v3.18.4 |
On-Premise |
ipvs |
subctl v0.9.0 |
| 1.19.1 |
1.19.1 |
ok (pod cidr: 10.44.0.0/16; 10.144.0.0/16) |
ok (service cidr: 10.45.0.0/16; 10.145.0.0/16) |
calicoctl v3.18.4 |
On-Premise |
iptables |
subctl v0.9.0 |
| TKE 独立集群1.18.4 |
TKE 独立集群1.18.4 |
ok (pod cidr: 10.1.1.0/24; 10.0.1.0/24) |
ok (service cidr: 10.55.0.0/24; 10.155.0.0/24) |
VPC-CNI |
Cloud |
ipvs |
subctl v0.9.0; |
| TKE 独立集群1.18.4 |
TKE 独立集群1.18.4 |
ok (pod cidr: 10.0.0.0/24; 10.0.1.0/24) |
ok (service cidr: 10.45.0.0/24; 10.145.0.0/24) |
VPC-CNI |
Cloud |
iptables |
subctl v0.9.0; |
| TKE 独立集群1.18.4 |
GKE 独立集群1.19.10-gke.1600 |
ok (pod cidr: 10.0.0.0/24; 10.4.0.0/14) |
ok (service cidr: 10.45.0.0/24; 10.8.0.0/20) |
VPC-CNI; VPC-native traffic routing |
Cloud |
iptables |
subctl v0.9.0 |
Reference:
Problem:
After multi clusters registered to a unfied control panel, the pod IP reachability between pods and services across clusters are required for HA and distributed application scheduling case, the submariner is the best choice against
cilium mashandistio multi cluster mesh, since it has little dependence and better align withregistered clusterdeploy topologyScope:
brokerandagentcomponents by CRCase result:
Reference: