Kubernetes pod vs container: Multi-container pods and container communication
Containers are often intended to solve a single, narrowly defined problem, such as a microservice, but in the real world, problems require multiple containers for a complete solution. In this article, we're going to talk about combining multiple containers into a single Kubernetes Pod, and what it means for inter-container communication.
What is a Kubernetes Pod?
Let's start by explaining what a Pod is in the first place. A Pod is is the smallest unit that can be deployed and managed by Kubernetes. In other words, if you need to run a single container in Kubernetes, then you need to create a Pod for that container. At the same time, a Pod can contain more than one container, usually because these containers are relatively tightly coupled. How tightly coupled? Well, think of it this way: the containers in a pod represent processes that would have run on the same server in a pre-container world.
And that makes sense, because in many respects, a Pod acts like a single server. For example, each container can access the other containers in the pod as different ports on localhost.
Why does Kubernetes use a Pod as the smallest deployable unit, and not a single container?
While it would seem simpler to just deploy a single container directly, there are good reasons to add a layer of abstraction represented by the Pod. A container is an existing entity, which refers to a specific thing. That specific thing might be a Docker container, but it might also be a rkt container, or a virtual machine (VM) managed by Virtlet. Each of these has different requirements.
What's more, Kubernetes needs additional information for container management, such as a restart policy, which defines what to do with a container when it terminates, or a liveness probe, which defines an action to detect if a process in a container is still alive from the application's perspective, such as a web server responding to HTTP requests.
Instead of overloading the existing "thing" with additional properties, Kubernetes architects have decided to use a new entity, the Pod, that logically contains (wraps) one or more containers that should be managed as a single entity.
Why does Kubernetes allow more than one container in a Pod?
Containers in a Pod run on a "logical host"; they use the same network namespace (in other words, the same IP address and port space), and the same IPC namespace. They can also use shared volumes. These properties make it possible for these containers to efficiently communicate, ensuring data locality. Also, Pods enable you to manage several tightly coupled application containers as a single unit.
So if an application needs several containers running on the same host, why not just make a single container with everything you need? Well first, you're likely to violate the "one process per container" principle. This is important because with multiple processes in the same container, it is harder to troubleshoot the container because logs from different processes will be mixed together, and it is harder to manage the processes lifecycle, for example to take care of "zombie" processes when their parent process dies. Second, using several containers for an application is simpler and more transparent, and enables decoupling software dependencies. Also, more granular containers can be reused between teams.
Use Cases for Multi-Container Pods
The primary purpose of a multi-container Pod is to support co-located, co-managed helper processes for a primary application. There are some general patterns for using helper processes in Pods:
Sidecar Containers Examples
Sidecar containers "help" the main container. Some examples include log or data change watchers, monitoring adapters, and so on. A log watcher, for example, can be built once by a different team and reused across different applications. Another example of a sidecar container is a file or data loader that generates data for the main container.
Proxies, bridges, and adapters
Proxies, bridges, and adapters connect the main container with the external world. For example, Apache HTTP server or nginx can serve static files. It can also act as a reverse proxy to a web application in the main container to log and limit HTTP requests. Another example is a helper container that re-routes requests from the main container to the external world. This makes it possible for the main container to connect to localhost to access, for example, an external database, but without any service discovery.
While you can host a multi-tier application (such as WordPress) in a single Pod, the recommended way is to use separate Pods for each tier, for the simple reason that you can scale tiers up independently and distribute them across cluster nodes.
Communication between containers in a Pod
Having multiple containers in a single Pod makes it relatively straightforward for them to communicate with each other. They can do this using several different methods.
Shared volumes in a K8s Pod
In Kubernetes, you can use a shared Kubernetes Volume as a simple and efficient way to share data between containers in a Pod. For most cases, it is sufficient to use a directory on the host that is shared with all containers within a Pod.
Kubernetes Volumes enables data to survive container restarts, but these volumes have the same lifetime as the Pod. That means that the volume (and the data it holds) exists exactly as long as that Pod exists. If that Pod is deleted for any reason, even if an identical replacement is created, the shared Volume is also destroyed and created anew.
A standard use case for a K8 multi-container Pod with a shared Volume is when one container writes logs or other files to the shared directory, and the other container reads from the shared directory. For example, we can create a Pod like so:
apiVersion: v1
kind: Pod
metadata:
name: mc1
spec:
volumes:
- name: html
emptyDir: {}
containers:
- name: 1st
image: nginx
volumeMounts:
- name: html
mountPath: /usr/share/nginx/html
- name: 2nd
image: debian
volumeMounts:
- name: html
mountPath: /html
command: ["/bin/sh", "-c"]
args:
- while true; do
date >> /html/index.html;
sleep 1;
doneIn this example, we define a volume named html. Its type is emptyDir, which means that the volume is first created when a Pod is assigned to a node, and exists as long as that Pod is running on that node. As the name says, it is initially empty. The 1st container runs nginx server and has the shared volume mounted to the directory /usr/share/nginx/html. The 2nd container uses the Debian image and has the shared volume mounted to the directory /html. Every second, the 2nd container adds the current date and time into the index.html file, which is located in the shared volume. When the user makes an HTTP request to the Pod, the Nginx server reads this file and transfers it back to the user in response to the request.
You can check that the pod is working either by exposing the nginx port and accessing it using your browser, or by checking the shared directory directly in the containers:
$ kubectl exec mc1 -c 1st -- /bin/cat /usr/share/nginx/html/index.html
...
Fri Aug 25 18:36:06 UTC 2017
$ kubectl exec mc1 -c 2nd -- /bin/cat /html/index.html
...
Fri Aug 25 18:36:06 UTC 2017
Fri Aug 25 18:36:07 UTC 2017Inter-process communications (IPC)
Containers in a Pod share the same IPC namespace, which means they can also communicate with each other using standard inter-process communications such as SystemV semaphores or POSIX shared memory.
In the following example, we define a Pod with two containers. We use the same Docker image for both. The first container, producer, creates a standard Linux message queue, writes a number of random messages, and then writes a special exit message. The second container, consumer, opens that same message queue for reading and reads messages until it receives the exit message. We also set the restart policy to 'Never', so the Pod stops after termination of both containers.
apiVersion: v1
kind: Pod
metadata:
name: mc2
spec:
containers:
- name: producer
image: allingeek/ch6_ipc
command: ["./ipc", "-producer"]
- name: consumer
image: allingeek/ch6_ipc
command: ["./ipc", "-consumer"]
restartPolicy: NeverTo check this out, create the pod using kubectl create and watch the Pod status:
$ kubectl get pods --show-all -w
NAME READY STATUS RESTARTS AGE
mc2 0/2 Pending 0 0s
mc2 0/2 ContainerCreating 0 0s
mc2 0/2 Completed 0 29Now you can check logs for each container and verify that the 2nd container received all messages from the 1st container, including the exit message:
$ kubectl logs mc2 -c producer
...
Produced: f4
Produced: 1d
Produced: 9e
Produced: 27
$ kubectl logs mc2 -c consumer
...
Consumed: f4
Consumed: 1d
Consumed: 9e
Consumed: 27
Consumed: done
There is one major problem with this Pod, however, and it has to do with how containers start up.
Container dependencies and startup order
Currently, all containers in a Pod are being started in parallel and there is no way to define that one container must be started after other container. For example, in the IPC example, there is a chance that the second container might finish starting before the first one has started and created the message queue. In this case, the second container will fail, because it expects that the message queue already exists.
In a cloud native environment, it's always better to plan for failures outside of your immediate control. For example, one way to to fix this issue would be to change the application to wait for the message queue to be created.
That said, you do have the opportunity to use Init Containers, which run before application containers. What's more, they run in the order you specify. For example, consider this example from the official docs:
apiVersion: v1
kind: Pod
metadata:
name: myapp-pod
labels:
app: myapp
spec:
containers:
- name: myapp-container
image: busybox:1.28
initContainers:
- name: init-myservice
image: busybox:1.28
- name: init-mydb
image: busybox:1.28In this case, Kubernetes runs init-myservice until it is successful, then init-mydb, and then the myapp-container application. You can make use of Init Containers to ensure that any groundwork your app containers need has been laid.
Inter-container network communication
Containers in a Pod are accessible via "localhost"; they use the same network namespace. Also, for containers, the observable host name is a Pod’s name. Because containers share the same IP address and port space, you should use different ports in containers for incoming connections. In other words, applications in a Pod must coordinate their usage of ports.
In the following example, we will create a multi-container Pod where nginx in one container works as a reverse proxy for a simple web application running in the second container.
Step 1. Create a ConfigMap with the nginx configuration file. Incoming HTTP requests to port 80 will be forwarded to port 5000 on localhost:
apiVersion: v1
kind: ConfigMap
metadata:
name: mc3-nginx-conf
data:
nginx.conf: |-
user nginx;
worker_processes 1;
error_log /var/log/nginx/error.log warn;
pid /var/run/nginx.pid;
events {
worker_connections 1024;
}
http {
include /etc/nginx/mime.types;
default_type application/octet-stream;
sendfile on;
keepalive_timeout 65;
upstream webapp {
server 127.0.0.1:5000;
}
server {
listen 80;
location / {
proxy_pass http://webapp;
proxy_redirect off;
}
}
}Step 2. Create a multi-container Pod with the simple web app and nginx in separate containers. Note that for the Pod, we define only nginx port 80. Port 5000 will not be accessible outside of the Pod.
apiVersion: v1
kind: Pod
metadata:
name: mc3
labels:
app: mc3
spec:
containers:
- name: webapp
image: training/webapp
- name: nginx
image: nginx:alpine
ports:
- containerPort: 80
volumeMounts:
- name: nginx-proxy-config
mountPath: /etc/nginx/nginx.conf
subPath: nginx.conf
volumes:
- name: nginx-proxy-config
configMap:
name: mc3-nginx-confStep 3. Expose the Pod using the NodePort service:
$ kubectl expose pod mc3 --type=NodePort --port=80
service "mc3" exposedStep 4. Identify port on the node that is forwarded to the Pod:
$ kubectl describe service mc3
...
NodePort: <unset> 31418/TCP
...Now you can use your browser (or curl) to navigate to your node’s port to access the web application through reverse proxy, as in:
This request will then be forwarded to port 5000 of the webapp container.
Exposing multiple containers in a Pod
While this example shows how to use a single container to access other containers in the pod, it's quite common for several containers in a Pod to listen on different ports -- all of which need to be exposed. To make this happen, you can either create a single service with multiple exposed ports, or you can create a single service for every poirt you're trying to expose.
Where to go from here
By creating pods, Kubernetes provides a great deal of flexibility for orchestrating how containers behave, and how they communicate with each other. They can share file volumes, they can communicate over the network, and they can even communicate using IPC.
That's just the beginning, of course. Interested in seeing what else you can do and getting more hands-on with Kubernetes? Check out our Cloud Native Operations Bootcamp (CN251) - which covers most of the topics for the Certified Kubernetes Administrator (CKA) exam. Hope to see you there!
FAQ
What are containerized applications?
Containerized applications are applications that don't run directly on the host machine, but instead have been encapsulated into containers, which then run in a container engine. In many cases, a containerized application is also broken down into multiple pieces, each of which lives in its own container.
Containerized applications are distributed as container images, which act as templates that can be instantiated as running containers.
What are Kubernetes containers?
Strictly speaking, there is no such thing as a K8 container. Instead, Kubernetes works with the concept of pods, with pods being defined as consisting of one or more containers.
What is the difference between containers vs pods?
While containers are the smallest unit to be managed in a containerized application, Kubernetes doesn't manage containers directly. Instead, Kubernetes manages pods, each of which can itself include one or more containers. When Kubernetes replicates or destroys a pod, the action includes all of the containers included in that pod.
