Use reflector.Run instead of reflector.ListAndWatch

[MaxFedotov]

We recently found out a problem in capsule-proxy when switching to rolebindings reflector to show tenant namespaces. If there are some issues with etcd (for example reelection is triggered) and reflector is unable to watch rolebindings from ApiServer

W1111 15:17:54.839976       1 reflector.go:424] storage/cacher.go:/rolebindings: watch of *rbac.RoleBinding ended with: Internal error occurred: etcdserver: no leader

it doesn't reconnect and watch new rolebindings until resync is triggered by rolebindings-resync-period.

In order to fix it will we had to use reflector.Run function (https://github.com/kubernetes/client-go/blob/v0.22.3/tools/cache/reflector.go#L218), which wraps reflector.ListAndWatch into backoff and will restart reflector in case of error returned

[MaxFedotov]

Just reproduced this behavior on my local laptop with Kind.

1. Create kind cluster, install capsule and capsule-proxy
2. Create tenant and generate tenant owner kubeconfig
3. Using tenant owner kubeconfig create namespace

kubectl --kubeconfig=hack/alice.kubeconfig create ns oil-foo  
namespace/oil-foo created

4. It will be visible

 kubectl --kubeconfig=hack/alice.kubeconfig get ns
NAME      STATUS   AGE
oil-foo   Active   7s

5. exec into capsule-control-plane pod and send SIGHUP to etcd, which will cause it to restart

root@capsule-control-plane:/# kill -HUP $(pidof etcd)

6. After etcd will be restarted, create new namespace

kubectl --kubeconfig=hack/alice.kubeconfig create ns oil-baz 
namespace/oil-baz created

7. And try to get them one more time

 kubectl --kubeconfig=hack/alice.kubeconfig get ns
NAME      STATUS   AGE
oil-foo   Active   7s

it will be missing

[MaxFedotov]

and now the same thing, but using reflector.Run instead of reflector.ListAndWatch in this func (https://github.com/clastix/capsule-proxy/blob/master/internal/controllers/role_bindings.go#L95)

1. Using tenant owner kubeconfig create namespace

kubectl --kubeconfig=hack/alice.kubeconfig create ns oil-foo  
namespace/oil-foo created

2. It will be visible

kubectl --kubeconfig=hack/alice.kubeconfig get ns
NAME      STATUS   AGE
oil-foo   Active   2s

3. exec into capsule-control-plane pod and send SIGHUP to etcd, which will cause it to restart

root@capsule-control-plane:/# kill -HUP $(pidof etcd)

4. After etcd will be restarted, create new namespace

kubectl --kubeconfig=hack/alice.kubeconfig create ns oil-baz 
namespace/oil-baz created

5. And try to get them one more time

kubectl --kubeconfig=hack/alice.kubeconfig get ns            
NAME      STATUS   AGE
oil-baz   Active   2s
oil-foo   Active   100s

it will be visible :)