This is a part of the series of posts on Getting memcache up and running on Kubernetes which explained how to create your first cluster and Installing memcache with Kubernetes which installed some memcache instances on your cluster and Exposing a memcache loadbalancer which makes it available externally. You’ll also have created a test app in your favorite node environment, and tested the memcache integration as described in Creating a test app for memcache on Kubernetes.  You also should have got the app running on Kubernetes as described in Getting a simple app running on Kubernetes
What we discovered was that although memcache works properly with each platform, the cache results are not reliably available across platforms, so we need to implement a router to synchronize them.

Helm install

Luckily there is already a Helm chart to install mcrouter. We used a similar process to install memcached in the first place as described in Installing memcache with Kubernetes. (which you should refer to if you don’t have Helm installed any longer).
First though, we’ll get rid of the memcached instances we already have.

$ helm delete mycache

and the exposed loadbalancer service
$ kubectl delete service mc
An now install the mcrouter service

$ helm install stable/mcrouter –name=mcrouter>

Expose the service

$ kubectl expose service mcrouter-memcached –port=11211 –target-port=11211 –name=mcrouter –type=LoadBalancer


You should now have these pods
$ kubectl get pods
mcdemo-566dfddd79-9d446 1/1 Running       0 16m
mcdemo-566dfddd79-pjw7p 1/1 Running       0 15m
mcrouter-mcrouter-4zc8l 1/1 Running       0 38m
mcrouter-mcrouter-f4v8x 1/1 Running       0 38m
mcrouter-mcrouter-phw6z 1/1 Running       0 38m
mcrouter-memcached-0    1/1 Running       0 38m
mcrouter-memcached-1    1/1 Running       0 37m
mcrouter-memcached-2    1/1 Running       0 37m

and these services

kubernetes         ClusterIP   <none>       443/TCP          3d
mcdemo-expose      LoadBalancer 8081:31143/TCP   3h
mcdemo-service     ClusterIP <none>       8081/TCP         4h
mcrouter           LoadBalancer 11211:32737/TCP 39m
mcrouter-mcrouter  ClusterIP    None          <none>       5000/TCP        42m
mcrouter-memcached ClusterIP    None          <none>       11211/TCP       42m


You’ll need to modify your secrets file with the new dns name (for inside the cluster) and the mc-demo loadbalancer external ip address (outside)
module.exports = ((ns) => {

    ns.memcached = {
        defExpires: 30 * 60 * 2,
        maxExpires: 30 * 60 * 24,
        c9: {
            host: 'xxxxxx:11211', // temporarly exposed for testing 
            silo: 'some value',
            verbose: true,
        ku: {
            host: 'mcrouter-memcached.default.svc.cluster.local:11211',
            silo: 'some value',
            verbose: true

    return ns;

Does it work?

What we are hoping now is that a search from Kubernetes will cause its cached result to be visible if the same result is done on the standalone node app.
time without cache
1132 ‘ms to complete’
timing with cache
hit: 25960109ae793d5f3e351a5fb946726963a35f7c
1 ‘ms to complete’
The same search using the standalone node App
hit: 05041f462435653b85a5bc386cfd72b50fb8e4fb
10 ‘ms to complete’
Wonderful, it works!

Next step

Let’s see if we can get Apps Script working with the same memcached router

Why not join our forum, follow the blog or follow me on twitter to ensure you get updates when they are available.