Download the oc client binary: https://mirror.openshift.com/pub/openshift-v4/clients/ocp/4.2.7/
Extract the client, for example:
tar -xvf openshift-client-linux-4.2.7.tar.gz
You can now move the client into your $PATH
, for example:
mv oc /home/$USER/bin
(you may need to create the directory)
or you can use the client with a relative path: ./oc
For the rest of the tutorial we will assume the oc
client is in your path.
Fork this repository (https://github.com/patrickdillon/openshift-intern-workshop/fork) and clone it to your machine
We will use a 4.2 OpenShift cluster from the Red Hat Product Demo System for this workshop.
You can either use your username and password:
oc login https://<CLUSTER_URL>:6443 -u <USERNAME>
or you can use authentication token. To get a token, go to OpenShift Console, login, click your username in the top right corner and click Copy Login Command.
The whole command will be copied and you only need to paste it in the terminal and run it.
To be able to install pre-commit package bellow, you will need a tool pip
dnf install python-pip
Pre-commit is a framework for managing and maintaining multi-language pre-commit hooks. The tool is useful when you are submitting changes as it helps you keep code clean and prevent potential issues which might be noticed during review or not at all.
To install pre-commit, run:
pip install --user pre-commit
Take a look at the file .pre-commit-config.yaml
which contains configuration for pre-commit used in this repository.
To install the pre-commit hooks to your git repository run
pre-commit install
Now, when you git commit
a set of checks will be run against your changes and errors will be reported or even fixed automatically.
We need to deploy our application now. To verify you are successfully logged into OpenShift, you can run the following command:
oc project
If there is no project you could use, create a new one:
oc new-project <YOUR_NAME>-intern-workshop
OpenShift/Kubernetes uses JSON and YAML format to describe the deployment artifacts. You can find all the manifests in YAML format in openshift/
directory
To deploy the whole application, you can pass a directory to the oc apply command:
oc apply -f openshift/
Go to the OpenShift Console and you should see a build running. Wait for the build to finish and for the deployment to proceed. Then try to access the route URL. You can get the URL under HOST/PORT
by:
$ oc get route
If you tried to access the application URL you were probably presented with Application is not available error. This could happen for various reasons, but the first thing we can check is whether our hostname and port in the Flask application are configured properly.
Look at the last line in app.py file - you'll see we load the port from environment variable or use port 5000
as a default. Also look at the deployment config in openshift/app.deploymentconfig.yaml and focus on 2 things:
- a field
containerPort
in containers section - an environment variabe
PORT
As you can see these do not match. As changing containerPort
would also require changing the service ports definition, we'll go for the envirnment variable.
Go ahead and change the 5000
to 8080
to match the exposed port in the openshift/app.deploymentconfig.yaml
file. We will then need to update the deployment config in the cluster by running
oc apply -f openshift/app.deploymentconfig.yaml
Once the application is redeployed, try to access the route URL again. You should get response like this:
{"msg":"Forbidden","status":403}
To get through the authentication you need to provide the URL with a secret in a query parameter, so try to append the following to your route URL:
?secret=secret
Internal server error - that does not look good - what have we missed? Let's investigate logs again - go to OpenShift Console > Workloads > Pods and view the pod logs.
You will see something like the following error among the log messages:
HTTP response body: b'{"kind":"Status","apiVersion":"v1","metadata":{},"status":"Failure","message":"pods is forbidden: User \\"system:serviceaccount:vpavlin-os-ws:default\\" cannot list pods in the namespace \\"vpavlin-os-ws\\": no RBAC policy matched","reason":"Forbidden","details":{"kind":"pods"},"code":403}\n'
Our application is trying to access OpenShift API without proper authorization (you can see authentication is ok - OpenShift recognized the provided service account, but it does not have the correct rights to access resources it is trying to access).
We need to add a correct role to our service account. You can see the error mentions service account (or SA) default- The best practice would be to create a separate SA for this use case, but let's just change the default one for now.
We will need to add a view role for the SA and we can do this by running the following command
oc adm policy add-role-to-user view -z default
If the command succeeded, you will be able to reload the app URL and get back a JSON response such as:
{"me":"openshift-intern-workshop-x-xxxxx","pods":["openshift-intern-workshop-x-xxxxx"],"routes":["openshift-intern-workshop"],"services":[]}
Let's look at secrets now. Secrets and Config Maps are resources used for app configuration. Our application uses one secret as well - openshift/app.secret.yaml. Take a look at it.
The interesting part is in section data
, but you cannot easily read it. The secret is obfuscated by base64 encoding to make it slightly harder to leak it by showing to someone. If we want to read it, we can copy the value and pass it through the base64 decoder
echo "c2VjcmV0" | base64 -d
So the actual value is secret
. Let's change it now! Pick a new secret and use it in following command
echo -n "<MYNEWSECRET>" | base64
Now copy the value and edit the secret in OpenShift
oc edit secret openshift-intern-workshop
Look at the file openshift/app.deploymentconfig.yaml
and try to find how the secret is used there.
As secrets and config maps are mainly used in environment variables (which cannot be changed dynamically at runtime from outside of the container), we need to re-deploy our application to pick up the new secret.
oc rollout latest openshift-intern-workshop
Once the deployment is finished, you will need to provide the new secret in the URL to be able to access the application by attaching ?secret=<MYNEWSECRET>
to the route URL.
Health checks are an important tool for the lifecycle management of an application. Readiness and liveness probes can be used to determine that a container is functioning properly.
A readiness probe is a check which verifies if your application is fully up and running and ready to accept requests. When readiness probes fail the container will not be assigned an ip address.
Liveness probes are subsequently used to repeatedly verify the application is up. If a liveness probe fails, the container will be restarted (based on the restart policy).
For our application, look at app.py and you will see a path called "health" which we can use in our health check.
Let's create a liveness probe.
Go OpenShift Console > Workloads > Deployment Configs > openshift-intern-workshop and click YAML.
You will see a section that looks like this:
[...]
containers:
- resources: {}
terminationMessagePath: /dev/termination-log
name: openshift-intern-workshop
env:
- name: PORT
value: '8080'
[...]
We need to add a livenessProbe
to our openshift-intern-workshop
container. Please note that the ordering of the fields below are not important but the livenessProbe
section has to indented correctly.
[...]
containers:
- resources: {}
terminationMessagePath: /dev/termination-log
name: openshift-intern-workshop
livenessProbe:
httpGet:
path: /health
port: 8080
initialDelaySeconds: 15
timeoutSeconds: 1
env:
- name: PORT
value: '8080'
[...]
Click Save and wait for a new deployment. If you click Pods, open the latest openshift-intern-workshop pod and select the openshift-intern-workshop
container, you should see the liveness probe registered on the left hand side.
Readiness probes can be created in the same way by replacing livenessProbe
with readinessProbe
.
Another important feature of OpenShift and Kubernetes is to make sure your application has enough resources, but at the same time does not consume more than an administrator allows. For this, we use resource requests and limits in the pod specification.
Limits ensure that your application does not consume too many resources and the OpenShift controller will kill the container if it does.
We have already edited Kubernetes objects using two methods: oc edit
and editing YAML through the console. We will change the limits of our application by editing the DeploymentConfig but let's do it in yet another way, the oc patch
command:
oc patch deploymentconfig openshift-intern-workshop -p '{"spec":{"template":{"spec":{"containers":[{"name":"openshift-intern-workshop", "resources": {"limits": {"memory": "600Mi", "cpu":"200m"}}}]}}}}'
This limits our application to 200
millicores and 600
megabytes of RAM.
Requests make sure your application has sufficient resources.
oc patch dc openshift-intern-workshop -p '{"spec":{"template":{"spec":{"containers":[{"name":"openshift-intern-workshop", "resources": {"requests": {"memory": "300Mi", "cpu":"100m"}}}]}}}}'
Our application now has between 300-600 megabytes of memory and 100-200 millicores of CPU. With oc describe pod
we can see that our pod now has a Quality of Service of burstable:
QoS Class: Burstable
Now that we have set our memory and CPU requests, we can attempt to scale our application.
As we are loading the information about running pods directly from OpenShift API, we can change the number of pods and see if the API response changes
oc scale --replicas=5 dc openshift-intern-workshop
Go to your app URL and reload it couple times - you will see the list of pods is longer now.
Now that the application is running in multiple replicas, we would like to see some load balancing. Look at the name in the me
field - take that value and use it in next command to kill that container
oc delete pod <CONTENT_OF_ME_FIELD>
Now if you reload your browser tab, you will see the value in me
field changed - i.e. the traffic goes to a different pod as the original one no longer exists.
Let's scale our deployment down again to make sure we don't disrupt the cluster and that redeployment does not take unnecessarily long time
oc scale --replicas=1 dc openshift-intern-workshop
Our application uses Source-To-Image (or S2I). S2I is a smart tool which makes it easy to build application container images. Look at the openshift/app.buildconfig.yaml to see how the S2I strategy is configured.
You can notice we need to provide 3 pieces of information
- Source image
- Source repository
- Output image
Source image is a container image which was designed for working with S2I - apart from other features it contains assemble
and run
scripts - you can see an example here: https://github.com/sclorg/s2i-python-container/blob/master/3.6/s2i/bin/assemble - which are used during build and start of the container.
Source repository is a git repository containing application in a language matching the one of a source container image, so that the tools in the source image know how to install the application.
Output image is a name of an image stream where the resulting container image will be pushed.
To be able to successfully build from your own repository, do not forget to change the build config source repository to your own!
When you develop your code you will need to rebuild the container image for your application. Our application was originally built and deployed from a git repository. To be able to quickly rebuild your new changes you might want to skip the step of pushing your code to a repository and then kicking off the build.
To do that, you can use (make sure you are in the root of the repository)
oc start-build openshift-intern-workshop --from-dir=. -F
Start build command will start new build in OpenShift and --from-dir
will collect contents of a given directory, compress it and send it to OpenShift as a context directory for the new build. Parameter -F
fill redirect logs from the build to the terminal, so that you can easily look at how the build progresses.
Once the build is finished, OpenShift will automatically redeploy our application - this happens based on triggers defined in openshift/app.deploymentconfig.yaml
Webhooks are a powerful automation feature provided by both OpenShift and Github. OpenShift will act as a receiver of a webhook request and Github will produce webhook calls when we push to the repository.
First go to OpenShift Console > Builds > Build Configs > openshift-intern-workshop > Configuration and Copy with Secret the Github Webhook URL.
Next go to your Github workshop repository and click Setting > Webhooks > Add webhook. Paste the copied URL in Payload URL, change Content type to application/json
and disable SSL verification and confirm by clicking Add webhook.
Create a new branch in your repository
git checkout -b feature/services
Look at the code in workshop/openshift_info.py
and try to implement a method similar to get_pods
, but instead of a list of Pod names, make it to return list of Service names
Solution
File workshop/openshift_info.py
def get_services(self):
services_api = self.oapi_client.resources.get(
kind='Service',
api_version='v1')
service_list = services_api.get(namespace=self.namespace)
return self._get_names(service_list)
You are now making changes to your code in a new branch, but the build config is pulling from master
. To make sure you build from the latest changes, we will need to add ref
to our build config.
oc edit bc openshift-intern-workshop
Find a section source
and in there find uri
, which should point to your repository fork (Make sure it has correct value too!). Add the following line right under the uri
field and make sure the indentation is the same
ref: feature/services
To verify the change, go to OpenShift Console > Builds > Build Configs > openshift-intern-workshop > Configuration and check the value of Source Ref:
You can now commit and push your changes
git commit -a -m "Add service list to API"
git push --set-upstream origin feature/services
Look at the OpenShift Console > Builds > Builds > openshift-intern-workshop > History - you will see a new build running, if your webhook is configured correctly. Wait for the build and following deployment to finish and reload your application - you should see a service listed there as well now.
Sometimes an application needs some persistency. The most classic example are databases - without a persistent volume all the data you store would be lost on container restart - and restarts happen a lot in a distributed cloud environment.
To simulate this situation, we have an endpoint in our app which stores a value in a file. First get the route of the app and store it in an environment variable
APP_URL=$(oc get route openshift-intern-workshop -o jsonpath='{.spec.host}')
Next try to query the /iam
endpoint
curl $APP_URL/iam
You will see a message: Could not find the 'iam' file
We need to set the value first by doing a POST request to the endpoint
curl -X POST $APP_URL/iam/<YOUR_NAME_HERE>
If this succeeded, you should get your name back when you do the GET request on the endpoint again
curl $APP_URL/iam
Now let's delete/restart the pod and see that the value is gone
POD=$(oc get pods | grep Running | awk '{print $1}')
oc delete pod $POD
Hit the endpoint again when the pod comes back up
curl $APP_URL/iam
As you can see, the value is gone. So let's make sure it gets properly persisted next time - let's add a persistent volume to our application. OpenShift uses something called dynamic provisioning to generate persistent volume based on persistent volume claims (or PVCs). Our task is only to create a PVC artifact and attach it to the pod and OpenShift will handle the rest.
Ideally you would do this by adding another YAML files to your git repository, but for the sake of simplicity, let's do it manually form the OpenShift Console. Go to the console > Workloads > Deployment Configs > openshift-intern-workshop > Actions > Add storage.
Click Create storage. Give your new PVC a name and size (e.g. 1 GB). Click Create. Then provide a mount path - if you look into app.py
file, you'll notice that the value submitted to the /iam
endpoint is stored in a file ./iam
. The full path to the file is /opt/app-root/src/iam
. As the /opt/app-root/src
directory contains our application, we will want to persist the file in a subdirectory. For that set the Mount Path to
/opt/app-root/src/data
and click Add.
We need to change the path in the source code as well - edit the app.py file and set the IAM_FILE
value to /opt/app-root/src/data/iam
- the line will now look like this:
IAM_FILE = "/opt/app-root/src/data/iam"
To get the change in we need to rebuild the container image - you can push the change to your repository, or use the build from a local dir - you have tried both before.
oc start-build openshift-intern-workshop --from-dir=. -F
Once the image is rebuilt and the application redeployed, we can send the POST
request again
curl -X POST $APP_URL/iam/<YOUR_NAME_HERE>
Then check the value is set properly
curl $APP_URL/iam
Delete (restart) the pod
POD=$(oc get pods | grep Running | awk '{print $1}')
oc delete pod $POD
and when it comes back up, see that the value is still there
curl $APP_URL/iam
Want to learn more? Go to http://learn.openshift.com/ and try more free workshops.