I am attempting to connect to my memcached instance via golang. Both my app and Memcached will be in Kubernetes.
I couldn't find a good way (or example) on how to dynamically discover memcached and ensure that it will update if the pods scale.
I can manually create a client with the IP but not sure the best way to dynamically create clients in K8s with Golang. Any help would be appreciated, below is my test function that I am using (if it helps).
func TestMemcached(t *testing.T) {
mc := memcache.New("10.40.1.18:11211", "10.40.2.16:11211")
mc.Set(&memcache.Item{Key: "foo", Value: []byte("my value"), Expiration: 1})
it, err := mc.Get("foo")
if err != nil {
t.Error("Error pulling from cache:" + err.Error())
}
fmt.Println("Key: " + string(it.Key) + "Value: " + string(it.Value))
//Sleep for a second and a half to let the key expire
time.Sleep(1500 * time.Millisecond)
it, err = mc.Get("foo")
if err == nil {
t.Error("Got a hit when I didn't expect it:" + err.Error())
}
}
kubectl describe svc memcached:
Name: memcached
Namespace: redfalcon
Labels: app=memcached
chart=memcached-2.3.1
heritage=Tiller
release=memcached
Annotations: <none>
Selector: app=memcached
Type: ClusterIP
IP: None
Port: memcache 11211/TCP
TargetPort: memcache/TCP
Endpoints: 10.40.1.23:11211,10.40.2.24:11211
Session Affinity: None
Events: <none>
You can expose your memcache instance(s) as a [Service](https://kubernetes.io/docs/concepts/services-networking/service/ in Kubernetes.
This way in the client you only specify the internal DNS entry for your service which would be something like <service-name>.<k8s-namespace>.svc.cluster.local
.
In your code you would need something like this:
mc := memcache.New("memcached.redfalcon.svc.cluster.local:11211")
You can also use ConfigMaps to specify this configuration for your application.
Hope it helps!
Ok, I figured it out! It had to do with telepresence (https://www.telepresence.io/) that I am using for local development. It gives me access to all K8s resources from my developmnet environment. Unfortunately there is a defect in Linux where it cannot use K8s FQDN for lookup.
https://github.com/telepresenceio/telepresence/issues/161
Fixing is is sort of a hack, I port forwarded my K8s pod and when running in local development mode I changed my URL to "localhost:port". This worked but it isn't really testing Memcached as K8s would use it.
When I deploy into K8s it works fine.