General deployment with Infoblox integration §
Prologue §
For simplicity let's assume that you operate two geographically distributed clusters you want to enable global load-balancing for. In this example, two local clusters will represent those two distributed clusters.
- Let's switch the context to the first cluster
export KUBECONFIG=eu-cluster
- Copy the default
values.yaml
from k8gb chart to any convenient location, e.g.
cp chart/k8gb/values.yaml ~/k8gb/eu-cluster.yaml
- Modify the example configuration. Important parameters described below:
-
dnsZone
- this zone will be delegated to theedgeDNS
in your environment. E.g.yourzone.edgedns.com
-
edgeDNSZone
- this zone will be automatically configured by k8gb to delegate todnsZone
and will make k8gb controlled nodes act as authoritative server for this zone. E.g.edgedns.com
-
edgeDNSServers
stable DNS servers in your environment that is controlled by edgeDNS provider e.g. Infoblox so k8gb instances will be able to talk to each other through automatically created DNS names -
clusterGeoTag
to geographically tag your cluster. We are operatingeu
cluster in this example -
extGslbClustersGeoTags
contains Geo tag of the cluster(s) to talk with when k8gb is deployed to multiple clusters. Imagine your second cluster isus
so we tag it accordingly -
infoblox.enabled: true
to enable automated zone delegation configuration at edgeDNS provider. You don't need it for local testing and can optionally be skipped. Meanwhile, in this section we will cover a fully operational end-to-end scenario. The other parameters do not need to be modified unless you want to do something special. E.g. to use images from private registry -
Export Infoblox related information in the shell.
-
export WAPI_USERNAME=<WAPI_USERNAME>
export WAPI_PASSWORD=<WAPI_PASSWORD>
- Create the Infoblox secret which is used by k8gb to configure edgeDNS by running:
kubectl create ns k8gb
make infoblox-secret
-
Expose associated k8gb CoreDNS service for DNS traffic on worker nodes.
Check this document for detailed information.
-
Let's deploy k8gb to the first cluster. Most of the helper commands are abstracted by GNU
make
. If you want to look under the hood please check theMakefile
. In general, standard Kubernetes/Helm commands are used. Point deployment mechanism to your customvalues.yaml
make deploy-gslb-operator VALUES_YAML=~/k8gb/eu-cluster.yaml
- It should deploy k8gb pretty quickly. Let's check the pod status
kubectl -n k8gb get pod
NAME READY STATUS RESTARTS AGE
k8gb-76cc56b55-t779s 1/1 Running 0 39s
k8gb-coredns-799984c646-qz88m 1/1 Running 0 41s
-
Deploy k8gb to the second cluster by repeating the same steps with the exception of:
- Switch context to 2nd cluster with
export KUBECONFIG=us-cluster
- Create another custom
values.yaml
withcp ~/k8gb/eu-cluster.yaml ~/k8gb/us-cluster.yaml
- Create another geo tag to enable cross cluster communication:
clusterGeoTag
becomesus
extGslbClustersGeoTags
becomeseu
- Run the installation pointing to new values file
make deploy-gslb-operator VALUES_YAML=~/k8gb/us-cluster.yaml
- Switch context to 2nd cluster with
-
When your 2nd cluster is ready by checking with
kubectl -n k8gb get pod
, we can proceed with the sample application installation -
We will use well known testing community app of podinfo
helm repo add podinfo https://stefanprodan.github.io/podinfo
kubectl create ns test-gslb
helm upgrade --install podinfo --namespace test-gslb --set ui.message="us" podinfo/podinfo
As you can see above we did set special geo tag message in podinfo configuration matching cluster geo tag. It is just for demonstration purposes.
- Check that podinfo is running
kubectl -n test-gslb get pod
NAME READY STATUS RESTARTS AGE
podinfo-5cfcdc9c45-jbg96 1/1 Running 0 2m18s
- Let's create Gslb CRD to enable global load balancing for this application. Notice the podinfo Service name
kubectl -n test-gslb get svc
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
podinfo ClusterIP 10.96.250.84 <none> 9898/TCP,9999/TCP 9m39s
- Create a custom resource
~/k8gb/podinfogslb.yaml
describingGslb
as per the sample below:
apiVersion: k8gb.absa.oss/v1beta1
kind: Gslb
metadata:
name: podinfo
namespace: test-gslb
spec:
ingress:
rules:
- host: podinfo.cloud.example.com
http:
paths:
- backend:
serviceName: podinfo # This should point to Service name of testing application
servicePort: http
path: /
strategy:
type: roundRobin # Use a round robin load balancing strategy, when deciding which downstream clusters to route clients too
- And apply the resource in the target app namespace
kubectl -n test-gslb apply -f podinfogslb.yaml
gslb.k8gb.absa.oss/podinfo created
- Check Gslb resource
kubectl -n test-gslb get gslb
NAME AGE
podinfo 39s
- Check Gslb resource status
kubectl -n test-gslb describe gslb
Name: podinfo
Namespace: test-gslb
Labels: <none>
Annotations: API Version: k8gb.absa.oss/v1beta1
Kind: Gslb
Metadata:
Creation Timestamp: 2020-06-24T22:51:09Z
Finalizers:
k8gb.absa.oss/finalizer
Generation: 1
Resource Version: 14197
Self Link: /apis/k8gb.absa.oss/v1beta1/namespaces/test-gslb/gslbs/podinfo
UID: 86d4121b-b870-434e-bd4d-fece681116f0
Spec:
Ingress:
Rules:
Host: podinfo.cloud.example.com
Http:
Paths:
Backend:
Service Name: podinfo
Service Port: http
Path: /
Strategy:
Type: roundRobin
Status:
Geo Tag: us
Healthy Records:
podinfo.cloud.example.com:
172.17.0.10
172.17.0.7
172.17.0.8
Service Health:
podinfo.cloud.example.com: Healthy
Events: <none>
-
In the output above you should see that Gslb detected the
Healthy
status of underlyingpodinfo
standard Kubernetes Service -
Check that internal k8gb DNS servers are responding accordingly on this cluster
- Pick one of the worker nodes to test with
k get nodes -o wide NAME STATUS ROLES AGE VERSION INTERNAL-IP EXTERNAL-IP OS-IMAGE KERNEL-VERSION CONTAINER-RUNTIME test-gslb2-control-plane Ready master 53m v1.17.0 172.17.0.9 <none> Ubuntu 19.10 4.19.76-linuxkit containerd://1.3.2 test-gslb2-worker Ready <none> 52m v1.17.0 172.17.0.8 <none> Ubuntu 19.10 4.19.76-linuxkit containerd://1.3.2 test-gslb2-worker2 Ready <none> 52m v1.17.0 172.17.0.7 <none> Ubuntu 19.10 4.19.76-linuxkit containerd://1.3.2 test-gslb2-worker3 Ready <none> 52m v1.17.0 172.17.0.10 <none> Ubuntu 19.10 4.19.76-linuxkit containerd://1.3.2
- Use
dig
to make a DNS query to itdig +short @172.17.0.10 podinfo.cloud.example.com 172.17.0.8 172.17.0.10 172.17.0.7
- One of your workers should already return DNS responses constructed by Gslb based on service health information
- If edgeDNS was configured you can query your standard infra DNS directly and it should return the same
dig +short podinfo.cloud.example.com 172.17.0.8 172.17.0.10 172.17.0.7
- Pick one of the worker nodes to test with
-
Now it's time to deploy this application to the first
eu
cluster. The steps and configuration are exactly the same. Just changingui.message
toeu
kubectl create ns test-gslb
helm upgrade --install podinfo --namespace test-gslb --set ui.message="eu" podinfo/podinfo
- Apply exactly the same Gslb definition
kubectl -n test-gslb apply -f podinfogslb.yaml
- Check the Gslb resource status.
k -n test-gslb describe gslb podinfo
Name: podinfo
Namespace: test-gslb
Labels: <none>
Annotations: API Version: k8gb.absa.oss/v1beta1
Kind: Gslb
Metadata:
Creation Timestamp: 2020-06-24T23:25:08Z
Finalizers:
k8gb.absa.oss/finalizer
Generation: 1
Resource Version: 23881
Self Link: /apis/k8gb.absa.oss/v1beta1/namespaces/test-gslb/gslbs/podinfo
UID: a5ab509b-5ea2-49d6-982e-4129a8410c3e
Spec:
Ingress:
Rules:
Host: podinfo.cloud.example.com
Http:
Paths:
Backend:
Service Name: podinfo
Service Port: http
Path: /
Strategy:
Type: roundRobin
Status:
Geo Tag: eu
Healthy Records:
podinfo.cloud.example.com:
172.17.0.3
172.17.0.5
172.17.0.6
172.17.0.8
172.17.0.10
172.17.0.7
Service Health:
podinfo.cloud.example.com: Healthy
Events: <none>
-
Ideally you should already see that
Healthy Records
ofpodinfo.cloud.example.com
return the records from both of the clusters. Otherwise, give it a couple of minutes to sync up. -
Now you can check the DNS responses the same way as before.
dig +short podinfo.cloud.example.com
172.17.0.8
172.17.0.5
172.17.0.10
172.17.0.7
172.17.0.6
172.17.0.3
- And for the final end-to-end test, we can use
curl
to query the application
curl -s podinfo.example.com|grep message
"message": "eu",
curl -s podinfo.example.com|grep message
"message": "us",
curl -s podinfo.example.com|grep message
"message": "us",
curl -s podinfo.example.com||grep message
"message": "eu",
- As you can see specially marked
podinfo
returns different geo tags showing us the Global Round Robin strategy is working as expected
Hope you enjoyed the ride!
If anything unclear or is going wrong, feel free to contact us at https://github.com/AbsaOSS/k8gb/issues. We will appreciate any feedback/bug report and Pull Requests are welcome.
For more advanced technical documentation and fully automated local installation steps, see below.