HeYe is a high-performance, high-scalability, ops-friendly, kubernetes-native loadbalancer.
HeYe provides traffic forward features, which handle traditional north-south traffic, as well as east-west traffic in datacenter efficiently.
The technical architecture of HeYe LB:
-
The data plane and the control plane are separated in the architecture, a clean architecture is realized. The data plane is borrowed from kubernetes, and the control plane is implemented by ourselves, which is very lightweight.
-
The forward traffic nodes are stateless and support horizontal scaling. The maximum number of nodes is 254, and the computing power and bandwidth are close to infinity.
-
Traffic forwarding is implemented through LVS. As we all know, LVS is an implementation of Layer 4 load balancing inside the Linux kernel. Layer 4 load balancing works on OSI Layer 4 (Transport Layer) and distributes requests to the servers at the transport layer without looking at the content of the packets, so it has the best performance.
-
General x86 server was liked without any special configuration or hardware.
-
The configuration of addition, deletion, modification and query supports hot update, which has no impact on the business. These can be done through yam and API, which is very simple and friendly
-
Supports 3 modes of health check on layer 7 HTTP headers, layer 4 tcp port and mysql service.
-
Supports multiple load balancing modes, such as RR, LC, WLC, WRR, source hash, etc.
Prepare
-
core switch supports network quality feature, such as cisco sla, h3c nqa
-
3 servers, minimal configuration(Strongly recommended):
-
8c, 32G mem, 100G disk
-
turnoff swap on all nodes,supports passwordless login beetween nodes
-
Installation
-
equivalent routing is configured on the switch and all traffic which forwards svc_cidr goes through three nodes
-
login node1
-
cd your_path/heyelb/install_cli
-
sh master_node.sh <node2_ip> <node3_ip> <node2_hostname> <node3_hostname> <svc_ip_cidr> <pod_ip_cidr> <etcd_endpoints>
for example: sh master_node.sh 10.1.1.2 10.1.1.3 k8s02 k8s03 10.1.1.4 10.11.0.0 10.12.0.0 "https://10.1.1.92:2379,https://10.1.1.93:2379,https://10.1.1.150:2379"
-
sh worker_node.sh <etcd_endpoints> <svc_ip_DNS> "cluster.local." "/kube-centos/network"
-
docker pull heyelb:v6
-
choose anyone node and run heyelb:
-
mkdir -p /workdir/servers/ && mkdir -p /workdir/etcd/ && mkdir -p /workdir/kubeconfig/
-
cd /workdir/servers/ && touch upstream.conf
-
cp your_path/admin.kubeconfig /workdir/kubeconfig/ && cd /workdir/kubeconfig/ && mv admin.kubeconfig kubeconfig.yaml
-
docker run -v /workdir/servers/:/export/home/yeepine-1.0/conf/servers/ -v /workdir/etcd:/default.etcd -v /workdir/kubeconfig:/workdir/kubeconfig -d heyelb:v6 /bin/bash
-
Getting started
-
edit a yaml file, Service.spec.ports.name and Endpoints.subnets.addresses.ports.name must be the same, Endpoints.metadata.annotations.calledSource and Endpoints.metadata.annotations.calledSource.healthCheckType are required.
-
Endpoints.metadata.annotations.calledSource.healthCheckType have 3 types, which supports only http, tcp and mysql.
ok, let's have a nice journey!
if you have any ideas or questions, please tell me([email protected]) or take a new issue. thanks